You are viewing a plain text version of this content. The canonical link for it is here.
Posted to general@incubator.apache.org by Alex Harui <ah...@adobe.com> on 2014/10/21 00:08:32 UTC
Convenience Binary Policy
Hi,
I’m wondering whether modifications to the set of bundled jars in a
convenience binary package can be made after release without voting.
And if not, I’m looking for any other quick-fix ideas for the following
scenario.
Flex has many different release packages. One is an SDK called FlexJS
0.0.2. Another is an Installer application. Most folks use the installer
to download Flex binary packages, unpack them and execute a script in the
package to download any dependencies.
The FlexJS install script was working great until about two weeks ago,
when the code the installer uses to fetch a jar from SourceForge stopped
working. It wasn’t a major problem because FlexJS is in ‘alpha’ stage and
only about 5 folks download it per day and they can go get the binary
package and use Ant to run the script and it will succeed. However, last
night, a community member realized that he was giving a talk on FlexJS on
Tuesday morning which could cause a rise in the number of folks who would
try to use the installer and subsequently fail. Any new vote plus mirror
propagation time will not be in time for the talk.
A workaround I tested was to add the jar from SourceForge to the binary
package. No other files are touched, although I suppose I should update
LICENSE and NOTICE. This works because the install script sees the jar
and skips trying to download it. I can take this modified binary package,
host it somewhere, change the installer’s config.xml that it fetches from
flex.a.o so that it will pick up this package instead of the one on dist
(actually, the mirrors) and the FlexJS 0.0.2 install will start working
again.
I know we can’t go messing around with source packages without a vote, but
what about binary packages? Is it against policy to do something like
this, and if so, can exceptions be made?
Thanks,
-Alex
Re: Convenience Binary Policy
Posted by Alex Harui <ah...@adobe.com>.
Let me see if I can tie all of the responses back into one thread.
Thanks Ross, for confirming that vote periods can be shorter. However,
there still isn’t enough time to get through vote + mirror latency before
Tuesday in San Francisco.
I think Ted just said I can update the binary package but I have to “label
it according to what it is”. I think that implies that it cannot be the
binary package for FlexJS 0.0.2, but why not? Is it because it isn’t a
true result of running the build script? Or because I should tweak
LICENSE and NOTICE in the binary package?
Ted also said that I could post a new installer. If I post an installer
that only offers to install this modified binary package, none of the
installer code needs to change other than where it picks up the list of
packages to offer. But that’s a source code change, so isn’t that like
advertising a nightly/development build to the public? Or can you offer
nightly binary packages, but not nightly source packages?
Another option for a modified installer is one that codes around the
SourceForge download problem. I’m investigating that now, but that will
also be a source code change.
Brane said that the Flex Community is used to voting on binary packages,
but I don’t think that’s true. We’re just trying to find out if there
really is any policy against modification of a binary package. I think
our PMC is more concerned about making new visitors to Apache and Flex
happy as long as we don’t break any rules.
-Alex
On 10/20/14, 10:01 PM, "Ross Gardler (MS OPEN TECH)"
<Ro...@microsoft.com> wrote:
>
>Ok. Well remember that the release vote period is a guideline. If this is
>such a trivial change maybe it would be acceptable to use a shorter vote
>period. As long as you have three +1 (meaning three people have verified
>the release) you would be good to go, without long debates about policy
>and intent ;-)
>
>Having said that it's always good to clarify things.
>
>-----Original Message-----
>From: Alex Harui [mailto:aharui@adobe.com]
>Sent: Monday, October 20, 2014 9:41 PM
>To: general@incubator.apache.org
>Subject: Re: Convenience Binary Policy
>
>Sorry, my last response crossed paths with this.
>
>We can and will make another release, but no, it was only 24 hours ago
>that we realized we might get a bump in installs from the talk on Tuesday
>and only 10 hours since I proved we could workaround the problem with a
>change to the binary package. No plan involving votes and mirrors would
>fix the problem in time. So I am looking for reasons why we can/can’t
>update a binary package in less time than the whole vote + mirrors
>latency.
>
>Thanks,
>-Alex
>
>On 10/20/14, 9:09 PM, "Ross Gardler (MS OPEN TECH)"
><Ro...@microsoft.com> wrote:
>
>>Regardless of whether you can/can't do this (others are commentating, I
>>won't add to that) - wouldn't it be easier to just build a release and
>>call a vote. My guess is that you spent more than three days from
>>identification of the problem to distribution and discussion here.
>>Remember, if you take the time to make a release nobody can veto it
>>(although if there are good community reasons to not release you'd be
>>expected to honor that).
>>
>>Ross
>>
>>-----Original Message-----
>>From: Alex Harui [mailto:aharui@adobe.com]
>>Sent: Monday, October 20, 2014 4:47 PM
>>To: general@incubator.apache.org
>>Subject: Re: Convenience Binary Policy
>>
>>
>>
>>On 10/20/14, 4:13 PM, "Ted Dunning" <te...@gmail.com> wrote:
>>
>>>On Mon, Oct 20, 2014 at 3:08 PM, Alex Harui <ah...@adobe.com> wrote:
>>>
>>>> I know we can’t go messing around with source packages without a
>>>>vote, but what about binary packages? Is it against policy to do
>>>>something like this, and if so, can exceptions be made?
>>>>
>>>
>>>I may not have followed this quite correctly, here is what I
>>>understood of the situation as you described it:
>>>
>>>1) there is a bug in the FlexJS distro, considered low priority due to
>>>sparse use
>>>
>>>2) you needed a quickly corrected binary distribution
>>>
>>>3) you created a correct distribution artifact and put it somewhere
>>>non-Apache
>>>
>>>4) you aren't claiming that the artifact you created is an Apache
>>>release and you are pointing some workshop participants at your release.
>>>
>>>I fail to see any problem whatsoever in what you did. You used Apache
>>>software to create a derived work which you are asking people to use
>>>in an instructional setting. As far as I can tell, the only claim you
>>>are making is that your artifact is FlexJS with a fix that should be
>>>incorporated upstream before long.
>>>
>>>What's the problem?
>>Well, the use of the Installer sort of implies that folks are getting
>>the same binary kit that was on dist/mirrors. So one of our PMC
>>members is objecting to this plan, even though the net result of what
>>ends up on the user’s disk is the same. We won’t be pointing just the
>>workshop participants at this modified binary, essentially everyone who
>>uses the installer to install FlexJS 0.0.2 will get it.
>>
>>This may not be a fair analogy, but suppose you bundled an external jar
>>in a binary distro and found out much later that the jar was corrupted
>>and needed a quick fix. Would you do what I just did and post a
>>corrected binary somewhere outside Apache and then update your
>>downloads page to point just the binary link there instead of the usual
>>dist/mirrors? For Flex, we don’t need to update our downloads page
>>because the binary on dist/mirrors works if you unpack it yourself and
>>run Ant, so the Installer makes it a bit different. No flex.a.o page
>>is going to point there, but the Installer app you downloaded from
>>flex.a.o will point there.
>>
>>Maybe that’s a better question: are their policies about where and to
>>what the binary links on a project’s download page can point? Can it
>>point outside the ASF to stuff that wasn’t generated at the same time
>>as the source that was voted on?
>>
>>-Alex
>>
>>
>>
>>---------------------------------------------------------------------
>>To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
>>For additional commands, e-mail: general-help@incubator.apache.org
>
>
>---------------------------------------------------------------------
>To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
>For additional commands, e-mail: general-help@incubator.apache.org
>?B�KKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKCB�
>?�?[��X��ܚX�K??K[XZ[?�?�[�\�[?][��X��ܚX�P?[��X�]?܋�\?X�?K�ܙ�B��܈?Y??]?[ۘ[?
>?��[X[�?�??K[XZ[?�?�[�\�[?Z?[???[��X�]?܋�\?X�?K�ܙ�B
RE: Convenience Binary Policy
Posted by "Ross Gardler (MS OPEN TECH)" <Ro...@microsoft.com>.
Ok. Well remember that the release vote period is a guideline. If this is such a trivial change maybe it would be acceptable to use a shorter vote period. As long as you have three +1 (meaning three people have verified the release) you would be good to go, without long debates about policy and intent ;-)
Having said that it's always good to clarify things.
-----Original Message-----
From: Alex Harui [mailto:aharui@adobe.com]
Sent: Monday, October 20, 2014 9:41 PM
To: general@incubator.apache.org
Subject: Re: Convenience Binary Policy
Sorry, my last response crossed paths with this.
We can and will make another release, but no, it was only 24 hours ago that we realized we might get a bump in installs from the talk on Tuesday and only 10 hours since I proved we could workaround the problem with a change to the binary package. No plan involving votes and mirrors would fix the problem in time. So I am looking for reasons why we can/can’t update a binary package in less time than the whole vote + mirrors latency.
Thanks,
-Alex
On 10/20/14, 9:09 PM, "Ross Gardler (MS OPEN TECH)"
<Ro...@microsoft.com> wrote:
>Regardless of whether you can/can't do this (others are commentating, I
>won't add to that) - wouldn't it be easier to just build a release and
>call a vote. My guess is that you spent more than three days from
>identification of the problem to distribution and discussion here.
>Remember, if you take the time to make a release nobody can veto it
>(although if there are good community reasons to not release you'd be
>expected to honor that).
>
>Ross
>
>-----Original Message-----
>From: Alex Harui [mailto:aharui@adobe.com]
>Sent: Monday, October 20, 2014 4:47 PM
>To: general@incubator.apache.org
>Subject: Re: Convenience Binary Policy
>
>
>
>On 10/20/14, 4:13 PM, "Ted Dunning" <te...@gmail.com> wrote:
>
>>On Mon, Oct 20, 2014 at 3:08 PM, Alex Harui <ah...@adobe.com> wrote:
>>
>>> I know we can’t go messing around with source packages without a
>>>vote, but what about binary packages? Is it against policy to do
>>>something like this, and if so, can exceptions be made?
>>>
>>
>>I may not have followed this quite correctly, here is what I
>>understood of the situation as you described it:
>>
>>1) there is a bug in the FlexJS distro, considered low priority due to
>>sparse use
>>
>>2) you needed a quickly corrected binary distribution
>>
>>3) you created a correct distribution artifact and put it somewhere
>>non-Apache
>>
>>4) you aren't claiming that the artifact you created is an Apache
>>release and you are pointing some workshop participants at your release.
>>
>>I fail to see any problem whatsoever in what you did. You used Apache
>>software to create a derived work which you are asking people to use
>>in an instructional setting. As far as I can tell, the only claim you
>>are making is that your artifact is FlexJS with a fix that should be
>>incorporated upstream before long.
>>
>>What's the problem?
>Well, the use of the Installer sort of implies that folks are getting
>the same binary kit that was on dist/mirrors. So one of our PMC
>members is objecting to this plan, even though the net result of what
>ends up on the user’s disk is the same. We won’t be pointing just the
>workshop participants at this modified binary, essentially everyone who
>uses the installer to install FlexJS 0.0.2 will get it.
>
>This may not be a fair analogy, but suppose you bundled an external jar
>in a binary distro and found out much later that the jar was corrupted
>and needed a quick fix. Would you do what I just did and post a
>corrected binary somewhere outside Apache and then update your
>downloads page to point just the binary link there instead of the usual
>dist/mirrors? For Flex, we don’t need to update our downloads page
>because the binary on dist/mirrors works if you unpack it yourself and
>run Ant, so the Installer makes it a bit different. No flex.a.o page
>is going to point there, but the Installer app you downloaded from
>flex.a.o will point there.
>
>Maybe that’s a better question: are their policies about where and to
>what the binary links on a project’s download page can point? Can it
>point outside the ASF to stuff that wasn’t generated at the same time
>as the source that was voted on?
>
>-Alex
>
>
>
>---------------------------------------------------------------------
>To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
>For additional commands, e-mail: general-help@incubator.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
For additional commands, e-mail: general-help@incubator.apache.org
Re: Convenience Binary Policy
Posted by Ted Dunning <te...@gmail.com>.
On Mon, Oct 20, 2014 at 9:40 PM, Alex Harui <ah...@adobe.com> wrote:
> So I am looking for reasons why we can/can’t
> update a binary package in less time than the whole vote + mirrors latency.
>
I think you can. Just label it according to what it is. You can even
link from the web site.
Re: Convenience Binary Policy
Posted by Alex Harui <ah...@adobe.com>.
Sorry, my last response crossed paths with this.
We can and will make another release, but no, it was only 24 hours ago
that we realized we might get a bump in installs from the talk on Tuesday
and only 10 hours since I proved we could workaround the problem with a
change to the binary package. No plan involving votes and mirrors would
fix the problem in time. So I am looking for reasons why we can/can’t
update a binary package in less time than the whole vote + mirrors latency.
Thanks,
-Alex
On 10/20/14, 9:09 PM, "Ross Gardler (MS OPEN TECH)"
<Ro...@microsoft.com> wrote:
>Regardless of whether you can/can't do this (others are commentating, I
>won't add to that) - wouldn't it be easier to just build a release and
>call a vote. My guess is that you spent more than three days from
>identification of the problem to distribution and discussion here.
>Remember, if you take the time to make a release nobody can veto it
>(although if there are good community reasons to not release you'd be
>expected to honor that).
>
>Ross
>
>-----Original Message-----
>From: Alex Harui [mailto:aharui@adobe.com]
>Sent: Monday, October 20, 2014 4:47 PM
>To: general@incubator.apache.org
>Subject: Re: Convenience Binary Policy
>
>
>
>On 10/20/14, 4:13 PM, "Ted Dunning" <te...@gmail.com> wrote:
>
>>On Mon, Oct 20, 2014 at 3:08 PM, Alex Harui <ah...@adobe.com> wrote:
>>
>>> I know we can’t go messing around with source packages without a
>>>vote, but what about binary packages? Is it against policy to do
>>>something like this, and if so, can exceptions be made?
>>>
>>
>>I may not have followed this quite correctly, here is what I understood
>>of the situation as you described it:
>>
>>1) there is a bug in the FlexJS distro, considered low priority due to
>>sparse use
>>
>>2) you needed a quickly corrected binary distribution
>>
>>3) you created a correct distribution artifact and put it somewhere
>>non-Apache
>>
>>4) you aren't claiming that the artifact you created is an Apache
>>release and you are pointing some workshop participants at your release.
>>
>>I fail to see any problem whatsoever in what you did. You used Apache
>>software to create a derived work which you are asking people to use in
>>an instructional setting. As far as I can tell, the only claim you are
>>making is that your artifact is FlexJS with a fix that should be
>>incorporated upstream before long.
>>
>>What's the problem?
>Well, the use of the Installer sort of implies that folks are getting the
>same binary kit that was on dist/mirrors. So one of our PMC members is
>objecting to this plan, even though the net result of what ends up on the
>user’s disk is the same. We won’t be pointing just the workshop
>participants at this modified binary, essentially everyone who uses the
>installer to install FlexJS 0.0.2 will get it.
>
>This may not be a fair analogy, but suppose you bundled an external jar
>in a binary distro and found out much later that the jar was corrupted
>and needed a quick fix. Would you do what I just did and post a
>corrected binary somewhere outside Apache and then update your downloads
>page to point just the binary link there instead of the usual
>dist/mirrors? For Flex, we don’t need to update our downloads page
>because the binary on dist/mirrors works if you unpack it yourself and
>run Ant, so the Installer makes it a bit different. No flex.a.o page is
>going to point there, but the Installer app you downloaded from flex.a.o
>will point there.
>
>Maybe that’s a better question: are their policies about where and to
>what the binary links on a project’s download page can point? Can it
>point outside the ASF to stuff that wasn’t generated at the same time as
>the source that was voted on?
>
>-Alex
>
>
>
>---------------------------------------------------------------------
>To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
>For additional commands, e-mail: general-help@incubator.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
For additional commands, e-mail: general-help@incubator.apache.org
RE: Convenience Binary Policy
Posted by "Ross Gardler (MS OPEN TECH)" <Ro...@microsoft.com>.
Regardless of whether you can/can't do this (others are commentating, I won't add to that) - wouldn't it be easier to just build a release and call a vote. My guess is that you spent more than three days from identification of the problem to distribution and discussion here. Remember, if you take the time to make a release nobody can veto it (although if there are good community reasons to not release you'd be expected to honor that).
Ross
-----Original Message-----
From: Alex Harui [mailto:aharui@adobe.com]
Sent: Monday, October 20, 2014 4:47 PM
To: general@incubator.apache.org
Subject: Re: Convenience Binary Policy
On 10/20/14, 4:13 PM, "Ted Dunning" <te...@gmail.com> wrote:
>On Mon, Oct 20, 2014 at 3:08 PM, Alex Harui <ah...@adobe.com> wrote:
>
>> I know we can’t go messing around with source packages without a
>>vote, but what about binary packages? Is it against policy to do
>>something like this, and if so, can exceptions be made?
>>
>
>I may not have followed this quite correctly, here is what I understood
>of the situation as you described it:
>
>1) there is a bug in the FlexJS distro, considered low priority due to
>sparse use
>
>2) you needed a quickly corrected binary distribution
>
>3) you created a correct distribution artifact and put it somewhere
>non-Apache
>
>4) you aren't claiming that the artifact you created is an Apache
>release and you are pointing some workshop participants at your release.
>
>I fail to see any problem whatsoever in what you did. You used Apache
>software to create a derived work which you are asking people to use in
>an instructional setting. As far as I can tell, the only claim you are
>making is that your artifact is FlexJS with a fix that should be
>incorporated upstream before long.
>
>What's the problem?
Well, the use of the Installer sort of implies that folks are getting the same binary kit that was on dist/mirrors. So one of our PMC members is objecting to this plan, even though the net result of what ends up on the user’s disk is the same. We won’t be pointing just the workshop participants at this modified binary, essentially everyone who uses the installer to install FlexJS 0.0.2 will get it.
This may not be a fair analogy, but suppose you bundled an external jar in a binary distro and found out much later that the jar was corrupted and needed a quick fix. Would you do what I just did and post a corrected binary somewhere outside Apache and then update your downloads page to point just the binary link there instead of the usual dist/mirrors? For Flex, we don’t need to update our downloads page because the binary on dist/mirrors works if you unpack it yourself and run Ant, so the Installer makes it a bit different. No flex.a.o page is going to point there, but the Installer app you downloaded from flex.a.o will point there.
Maybe that’s a better question: are their policies about where and to what the binary links on a project’s download page can point? Can it point outside the ASF to stuff that wasn’t generated at the same time as the source that was voted on?
-Alex
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
For additional commands, e-mail: general-help@incubator.apache.org
Re: Convenience Binary Policy
Posted by Alex Harui <ah...@adobe.com>.
On 10/20/14, 4:13 PM, "Ted Dunning" <te...@gmail.com> wrote:
>On Mon, Oct 20, 2014 at 3:08 PM, Alex Harui <ah...@adobe.com> wrote:
>
>> I know we can’t go messing around with source packages without a vote,
>>but
>> what about binary packages? Is it against policy to do something like
>> this, and if so, can exceptions be made?
>>
>
>I may not have followed this quite correctly, here is what I understood of
>the situation as you described it:
>
>1) there is a bug in the FlexJS distro, considered low priority due to
>sparse use
>
>2) you needed a quickly corrected binary distribution
>
>3) you created a correct distribution artifact and put it somewhere
>non-Apache
>
>4) you aren't claiming that the artifact you created is an Apache release
>and you are pointing some workshop participants at your release.
>
>I fail to see any problem whatsoever in what you did. You used Apache
>software to create a derived work which you are asking people to use in an
>instructional setting. As far as I can tell, the only claim you are
>making
>is that your artifact is FlexJS with a fix that should be incorporated
>upstream before long.
>
>What's the problem?
Well, the use of the Installer sort of implies that folks are getting the
same binary kit that was on dist/mirrors. So one of our PMC members is
objecting to this plan, even though the net result of what ends up on the
user’s disk is the same. We won’t be pointing just the workshop
participants at this modified binary, essentially everyone who uses the
installer to install FlexJS 0.0.2 will get it.
This may not be a fair analogy, but suppose you bundled an external jar in
a binary distro and found out much later that the jar was corrupted and
needed a quick fix. Would you do what I just did and post a corrected
binary somewhere outside Apache and then update your downloads page to
point just the binary link there instead of the usual dist/mirrors? For
Flex, we don’t need to update our downloads page because the binary on
dist/mirrors works if you unpack it yourself and run Ant, so the Installer
makes it a bit different. No flex.a.o page is going to point there, but
the Installer app you downloaded from flex.a.o will point there.
Maybe that’s a better question: are their policies about where and to what
the binary links on a project’s download page can point? Can it point
outside the ASF to stuff that wasn’t generated at the same time as the
source that was voted on?
-Alex
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
For additional commands, e-mail: general-help@incubator.apache.org
Re: Convenience Binary Policy
Posted by jan i <ja...@apache.org>.
On 21 October 2014 07:26, Roman Shaposhnik <ro...@shaposhnik.org> wrote:
> On Mon, Oct 20, 2014 at 9:45 PM, Branko Čibej <br...@apache.org> wrote:
> > On 21.10.2014 06:34, Alex Harui wrote:
> >> What is the piece I’m missing that says we have to vote to update the
> >> binary package?
> >
> > Apparently the Flex community believes that convenience binaries need
> > votes. They don't, but aside from that, if you guys are already voting
> > on binary packages, it makes perfect sense to vote for your fixed
> > version, if only to keep people happy.
> >
> >
> > -- Brane
> >
> > P.S.: Why anyone would think voting on binaries makes any kind of sense
> > around here is, of course, a different question. I can't even begin to
> > count the number of times it's been pointed out that binaries are not
> > Apache releases.
>
I am sure you make among other Apache OpenOffice unhappy with that
statement, we have several projects where the binaries are more frequent
downloaded than the source code.
Also me wonders, why has there been such a big discussion/demand for a
digitial signing service (just released by infra), that makes absolutely no
sense if we only release source.
It might be our rules say we only release source, but I am sure many of our
users in e.g. Apache OpenOffice users dont care about the source, their
only concern is the binary, as a consequence we need to be carefull
whenever a binary carries the apache name.
of traffic I've seen around clarifying some finer (and not so fine)
> points of release/licensing implication it is about time we start
> an FAQ on that. Me thinks at least.
>
+1
rgds
jan i
>
> Thanks,
> Roman.
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
> For additional commands, e-mail: general-help@incubator.apache.org
>
>
Re: Convenience Binary Policy
Posted by Marvin Humphrey <ma...@rectangular.com>.
Hello, Harbs,
My two cents:
Regardless of policy, I don't see how it makes sense from a governance
perspective for individuals to take controversial independent action on
something so important as release publication. The mere fact that a dispute
exists should block such changes.
Can't the Flex PMC resolve this issue on that basis?
Marvin Humphrey
On Wed, Oct 22, 2014 at 12:01 AM, Harbs <ha...@gmail.com> wrote:
> Thanks for the clarification!
>
> I’m still not sure I understand. In plain English we seem to have these unresolved questions:
>
> 1) (Re)compiling convenience packages with modifications to binary
> dependencies after the release vote: Is that kosher or not?
> 2) If a binary dependency is added to a convenience package (instead of
> being downloaded at install time) which results in the need to modify the
> LICENSE / NOTICE file that’s accompanying the convenience package, does
> that require a new VOTE and release or not?
> 3) In our case, the binary dependency is JBurg which is used as part of the
> compiler process on the client’s machine. It’s not byte code that is
> compiled into the resultant binary. The literal interpretation of the
> text would lean towards saying that’s a no-no. I find it hard to believe
> that that’s the intent. Is there any way to better clarify this point?
>
> Justin, please correct me if I missed any points or misrepresented any of
> them.
>
> Thanks,
> Harbs
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
For additional commands, e-mail: general-help@incubator.apache.org
Re: Convenience Binary Policy
Posted by Harbs <ha...@gmail.com>.
Thanks for the clarification!
I’m still not sure I understand. In plain English we seem to have these unresolved questions:
1) (Re)compiling convenience packages with modifications to binary dependencies after the release vote: Is that kosher or not?
2) If a binary dependency is added to a convenience package (instead of being downloaded at install time) which results in the need to modify the LICENSE / NOTICE file that’s accompanying the convenience package, does that require a new VOTE and release or not?
3) In our case, the binary dependency is JBurg which is used as part of the compiler process on the client’s machine. It’s not byte code that is compiled into the resultant binary. The literal interpretation of the text would lean towards saying that’s a no-no. I find it hard to believe that that’s the intent. Is there any way to better clarify this point?
Justin, please correct me if I missed any points or misrepresented any of them.
Thanks,
Harbs
On Oct 21, 2014, at 9:35 PM, Marvin Humphrey <ma...@rectangular.com> wrote:
> On Tue, Oct 21, 2014 at 6:55 AM, Harbs <ha...@gmail.com> wrote:
>> The one thing I see missing from the proposed text is dependencies and
>> installers.
>>
>> Particularly this section:
>>
>> ### Compiled packages ### {#compiled-packages}
>>
>> The Apache Software Foundation produces open source software. All releases
>> are in the form of the source materials needed to make changes to the
>> software being released.
>>
>> As a convenience to users that might not have the appropriate tools to
>> build a compiled version of the source, binary/bytecode packages MAY be
>> distributed alongside official Apache releases. In all such cases, the
>> binary/bytecode package MUST have the same version number as the source
>> release and MUST only add binary/bytecode files that are the result of
>> compiling that version of the source code release.
>>
>> ---
>>
>> How do binary dependencies fit in?
>
> Thanks for the review!
>
> The sentence in question is taken nearly verbatim from the existing release
> policy document. The only changes are capitalizing "MUST" and swapping out
> "may only" for "MUST ONLY":
>
> http://www.apache.org/dev/release#what
>
> In all such cases, the binary/bytecode package must have the same version
> number as the source release and may only add binary/bytecode files that
> are the result of compiling that version of the source code release.
>
> This is in keeping with the goal of the initiative: "to clarify policy,
> NOT TO CHANGE IT."
>
> My interpretation of that passage is that a liberal view of the word
> "compiling" allows the bundling of binary dependencies. Sometimes "compiling"
> is taken to mean a compilation/translation phase which is distinct from
> linking, and other times "compiling" encompasses linking (either static or
> dynamic). The more expansive definition is consistent with how that
> policy has traditionally been applied.
>
> Marvin Humphrey
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
> For additional commands, e-mail: general-help@incubator.apache.org
Re: Convenience Binary Policy
Posted by Marvin Humphrey <ma...@rectangular.com>.
On Tue, Oct 21, 2014 at 6:55 AM, Harbs <ha...@gmail.com> wrote:
> The one thing I see missing from the proposed text is dependencies and
> installers.
>
> Particularly this section:
>
> ### Compiled packages ### {#compiled-packages}
>
> The Apache Software Foundation produces open source software. All releases
> are in the form of the source materials needed to make changes to the
> software being released.
>
> As a convenience to users that might not have the appropriate tools to
> build a compiled version of the source, binary/bytecode packages MAY be
> distributed alongside official Apache releases. In all such cases, the
> binary/bytecode package MUST have the same version number as the source
> release and MUST only add binary/bytecode files that are the result of
> compiling that version of the source code release.
>
> ---
>
> How do binary dependencies fit in?
Thanks for the review!
The sentence in question is taken nearly verbatim from the existing release
policy document. The only changes are capitalizing "MUST" and swapping out
"may only" for "MUST ONLY":
http://www.apache.org/dev/release#what
In all such cases, the binary/bytecode package must have the same version
number as the source release and may only add binary/bytecode files that
are the result of compiling that version of the source code release.
This is in keeping with the goal of the initiative: "to clarify policy,
NOT TO CHANGE IT."
My interpretation of that passage is that a liberal view of the word
"compiling" allows the bundling of binary dependencies. Sometimes "compiling"
is taken to mean a compilation/translation phase which is distinct from
linking, and other times "compiling" encompasses linking (either static or
dynamic). The more expansive definition is consistent with how that
policy has traditionally been applied.
Marvin Humphrey
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
For additional commands, e-mail: general-help@incubator.apache.org
RE: Convenience Binary Policy
Posted by "Dennis E. Hamilton" <de...@acm.org>.
<orcnote> in-line.
-----Original Message-----
From: brane@apache.org [mailto:brane@apache.org]
Sent: Thursday, October 23, 2014 01:47
To: general@incubator.apache.org
Subject: Re: Convenience Binary Policy
On 22.10.2014 03:02, Justin Mclean wrote:
> Hi,
>
>> Binary dependencies are, by definition, not released by the ASF; because
>> we release source code. Also, software that has dependencies that are
>> only available in binary form is not open-source, in my book.
> You may possibly be forgetting about Category B licensed dependancies. These may only be included in binary form in an Apache product. [1] They would be still be consider open source by most people :-) The source does exist it's just not able to be included in a Apache product due to the weak copy left provision the licences include.
I have trouble visualising how any ASF project could have /mandatory/
dependencies on anything from the B-list.
<orcnote>
For an indication of the kinds of non-ASF dependencies that go into
Apache OpenOffice binary distributions, take a look at this list:
<http://sourceforge.net/projects/oooextras.mirror/files/?source=directory>.
(To see the full names, mouse over the names in the Name column.)
There is a non-neglible potential for more Category B dependencies
in the future, now that LibreOffice releases are under the MPL 2.0.
</orcnote>
-- Brane
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
For additional commands, e-mail: general-help@incubator.apache.org
Re: Convenience Binary Policy
Posted by Andrea Pescetti <pe...@apache.org>.
On 23/10/2014 brane wrote:
> On 22.10.2014 03:02, Justin Mclean wrote:
>> You may possibly be forgetting about Category B licensed
>> dependancies. These may only be included in binary form in an
>> Apache product. ...
> I have trouble visualising how any ASF project could have
> /mandatory/ dependencies on anything from the B-list.
If I understand correctly, we are speaking about optional dependencies
rather than mandatory dependencies. But the distinction, especially if
you look at the information Dennis already discussed, might not always
be clear: for OpenOffice, one of those dependencies is the spell checker
(external library, Hunspell).
So: our source release can perfectly come without a spell checker,
builds fine and works. Technically this is an "optional" dependency and
everything is correct and in accordance with the ASF rules. Still, for a
user it would be quite problematic if we removed the spell checker from
the convenience binaries... So this is not mandatory but it's still a
"must have" for our final users, and given the OpenOffice audience it
totally makes sense that the project recommends a convenience binary
that already comes with some optional features included.
Regards,
Andrea.
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
For additional commands, e-mail: general-help@incubator.apache.org
Re: Convenience Binary Policy
Posted by Harbs <ha...@gmail.com>.
I understand the resolution as follows:
Releases are for source only. The source can have a dependency on JBurg which the user would download themselves if they are installing from source. I guess we should include instructions to install from Source, so that’s an option for users. But we’d include a link so they could install using “convenience packages” as well.
Any “convenience packages” would be hosted outside and not be part of the official release process. The installer which would install the SDk along with Falcon and JBurg would fall in that category and be hosted on Github. Since the installer is not an “official Apache release”, it could be modified on Github as much as we want with no voting required.
Harbs
On Oct 23, 2014, at 3:42 PM, Justin Mclean <ju...@classsoftware.com> wrote:
> Hi,
>
>> The easiest might be to host the Flex installer outside of Apache
>> Flex, as a separate project where its contributors are free to do
>> whatever.
>
> I may be missing something here but I'm not sure that solves any issues. It's not the installer than needed to be modified but what it what it was installing that needed to be modified ie the FlexJS release. It's the FlexJS release has tthe dependancy on JBerg not the installer.
>
> Thanks,
> Justin
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
> For additional commands, e-mail: general-help@incubator.apache.org
>
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
For additional commands, e-mail: general-help@incubator.apache.org
Re: Convenience Binary Policy
Posted by Justin Mclean <ju...@classsoftware.com>.
Hi,
> The easiest might be to host the Flex installer outside of Apache
> Flex, as a separate project where its contributors are free to do
> whatever.
I may be missing something here but I'm not sure that solves any issues. It's not the installer than needed to be modified but what it what it was installing that needed to be modified ie the FlexJS release. It's the FlexJS release has tthe dependancy on JBerg not the installer.
Thanks,
Justin
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
For additional commands, e-mail: general-help@incubator.apache.org
Re: Convenience Binary Policy
Posted by Justin Mclean <ju...@classsoftware.com>.
Hi,
> I just created “flex-extras” and we’ll put in some appropriate text both on the Github side and Apache Flex side. Hopefully this will prove to be a smooth resolution.
Also are you also sure it's possible that you can upload and distribute large binary releases on github? See [1] [2] which seem to say you can't, although may be possible via the newish release feature [3]
Justin
1. https://github.com/blog/1302-goodbye-uploads
2. https://github.com/blog/1533-new-file-size-limits
3. https://github.com/blog/1547-release-your-software
Re: Convenience Binary Policy
Posted by Justin Mclean <ju...@classsoftware.com>.
HI,
> BTW, I still have plans to create a separate business entity that, among
> other things, will distribute Flex and FlexJS binaries that are
> essentially how Adobe distributed SDKs.
Other people already have business entities set up and can do this if they got permission/license from Adobe to do so.
I've contacted Adobe several times about this and no one seems who know who's responsible / how it can be done / got no reply. Do you know the right person to contact?
Thanks,
Justin
Re: Convenience Binary Policy
Posted by Harbs <ha...@gmail.com>.
My thinking was to move any installers and the like over to Github and they would be completely removed from the release process. The releases would be source-only.
Putting up builds with AIR (and the like) rolled in could be a nice plus.
On Oct 23, 2014, at 8:22 PM, Erik de Bruin <er...@ixsoftware.nl> wrote:
> The way I understand it, the proposal is to kinda “fork” the installer on
> occasions where a non-Apache binary distribution solves a problem. Or maybe
> it would just be a way to host a binary that was created using the regular
> Git ‘develop’ branch.
>
> Did I get that right?
>
> EdB
>
>
>
> On Thu, Oct 23, 2014 at 7:14 PM, OmPrakash Muppirala <bi...@gmail.com>
> wrote:
>
>> So what exactly goes in this flex-extras repo?
>>
>> Thanks,
>> Om
>>
>> On Thu, Oct 23, 2014 at 9:45 AM, Alex Harui <ah...@adobe.com> wrote:
>>
>>> I would definitely recommend that the MD5 and ASC hash/sigs be co-located
>>> with the artifacts.
>>>
>>> BTW, I still have plans to create a separate business entity that, among
>>> other things, will distribute Flex and FlexJS binaries that are
>>> essentially how Adobe distributed SDKs: AIR/Flash will be packaged in
>> it.
>>> No install necessary. Just unzip and point your IDE at it.
>>>
>>> I have to get clearance from Adobe before I do it, and just have been too
>>> swamped to do it, but my hope is that it will cut down significantly on
>>> failed installs since, once you get a package that unzips there will not
>>> be anything else to download and potentially fail.
>>>
>>> -Alex
>>>
>>> On 10/23/14, 7:33 AM, "Kessler CTR Mark J" <ma...@usmc.mil>
>>> wrote:
>>>
>>>> The only reason I bring this up; some people verify the package has not
>>>> been modified if getting it from another source.
>>>>
>>>> -Mark
>>>>
>>>> -----Original Message-----
>>>> From: Harbs [mailto:harbs.lists@gmail.com]
>>>> Sent: Thursday, October 23, 2014 10:00 AM
>>>> To: dev@flex.apache.org
>>>> Subject: Re: Convenience Binary Policy
>>>>
>>>> I think that's up to us.
>>>>
>>>> On Oct 23, 2014, at 4:56 PM, Kessler CTR Mark J
>>>> <ma...@usmc.mil> wrote:
>>>>
>>>>> I like the idea so far. Would there still be hashes / signatures on
>>>>> the non-official releases?
>>>>>
>>>>> -Mark
>>>>
>>>
>>>
>>
>
>
>
> --
> Ix Multimedia Software
>
> Jan Luykenstraat 27
> 3521 VB Utrecht
>
> T. 06-51952295
> I. www.ixsoftware.nl
Re: Convenience Binary Policy
Posted by Erik de Bruin <er...@ixsoftware.nl>.
The way I understand it, the proposal is to kinda “fork” the installer on
occasions where a non-Apache binary distribution solves a problem. Or maybe
it would just be a way to host a binary that was created using the regular
Git ‘develop’ branch.
Did I get that right?
EdB
On Thu, Oct 23, 2014 at 7:14 PM, OmPrakash Muppirala <bi...@gmail.com>
wrote:
> So what exactly goes in this flex-extras repo?
>
> Thanks,
> Om
>
> On Thu, Oct 23, 2014 at 9:45 AM, Alex Harui <ah...@adobe.com> wrote:
>
> > I would definitely recommend that the MD5 and ASC hash/sigs be co-located
> > with the artifacts.
> >
> > BTW, I still have plans to create a separate business entity that, among
> > other things, will distribute Flex and FlexJS binaries that are
> > essentially how Adobe distributed SDKs: AIR/Flash will be packaged in
> it.
> > No install necessary. Just unzip and point your IDE at it.
> >
> > I have to get clearance from Adobe before I do it, and just have been too
> > swamped to do it, but my hope is that it will cut down significantly on
> > failed installs since, once you get a package that unzips there will not
> > be anything else to download and potentially fail.
> >
> > -Alex
> >
> > On 10/23/14, 7:33 AM, "Kessler CTR Mark J" <ma...@usmc.mil>
> > wrote:
> >
> > >The only reason I bring this up; some people verify the package has not
> > >been modified if getting it from another source.
> > >
> > >-Mark
> > >
> > >-----Original Message-----
> > >From: Harbs [mailto:harbs.lists@gmail.com]
> > >Sent: Thursday, October 23, 2014 10:00 AM
> > >To: dev@flex.apache.org
> > >Subject: Re: Convenience Binary Policy
> > >
> > >I think that's up to us.
> > >
> > >On Oct 23, 2014, at 4:56 PM, Kessler CTR Mark J
> > ><ma...@usmc.mil> wrote:
> > >
> > >> I like the idea so far. Would there still be hashes / signatures on
> > >>the non-official releases?
> > >>
> > >> -Mark
> > >
> >
> >
>
--
Ix Multimedia Software
Jan Luykenstraat 27
3521 VB Utrecht
T. 06-51952295
I. www.ixsoftware.nl
Re: Convenience Binary Policy
Posted by OmPrakash Muppirala <bi...@gmail.com>.
So what exactly goes in this flex-extras repo?
Thanks,
Om
On Thu, Oct 23, 2014 at 9:45 AM, Alex Harui <ah...@adobe.com> wrote:
> I would definitely recommend that the MD5 and ASC hash/sigs be co-located
> with the artifacts.
>
> BTW, I still have plans to create a separate business entity that, among
> other things, will distribute Flex and FlexJS binaries that are
> essentially how Adobe distributed SDKs: AIR/Flash will be packaged in it.
> No install necessary. Just unzip and point your IDE at it.
>
> I have to get clearance from Adobe before I do it, and just have been too
> swamped to do it, but my hope is that it will cut down significantly on
> failed installs since, once you get a package that unzips there will not
> be anything else to download and potentially fail.
>
> -Alex
>
> On 10/23/14, 7:33 AM, "Kessler CTR Mark J" <ma...@usmc.mil>
> wrote:
>
> >The only reason I bring this up; some people verify the package has not
> >been modified if getting it from another source.
> >
> >-Mark
> >
> >-----Original Message-----
> >From: Harbs [mailto:harbs.lists@gmail.com]
> >Sent: Thursday, October 23, 2014 10:00 AM
> >To: dev@flex.apache.org
> >Subject: Re: Convenience Binary Policy
> >
> >I think that's up to us.
> >
> >On Oct 23, 2014, at 4:56 PM, Kessler CTR Mark J
> ><ma...@usmc.mil> wrote:
> >
> >> I like the idea so far. Would there still be hashes / signatures on
> >>the non-official releases?
> >>
> >> -Mark
> >
>
>
Re: Convenience Binary Policy
Posted by Alex Harui <ah...@adobe.com>.
I would definitely recommend that the MD5 and ASC hash/sigs be co-located
with the artifacts.
BTW, I still have plans to create a separate business entity that, among
other things, will distribute Flex and FlexJS binaries that are
essentially how Adobe distributed SDKs: AIR/Flash will be packaged in it.
No install necessary. Just unzip and point your IDE at it.
I have to get clearance from Adobe before I do it, and just have been too
swamped to do it, but my hope is that it will cut down significantly on
failed installs since, once you get a package that unzips there will not
be anything else to download and potentially fail.
-Alex
On 10/23/14, 7:33 AM, "Kessler CTR Mark J" <ma...@usmc.mil>
wrote:
>The only reason I bring this up; some people verify the package has not
>been modified if getting it from another source.
>
>-Mark
>
>-----Original Message-----
>From: Harbs [mailto:harbs.lists@gmail.com]
>Sent: Thursday, October 23, 2014 10:00 AM
>To: dev@flex.apache.org
>Subject: Re: Convenience Binary Policy
>
>I think that's up to us.
>
>On Oct 23, 2014, at 4:56 PM, Kessler CTR Mark J
><ma...@usmc.mil> wrote:
>
>> I like the idea so far. Would there still be hashes / signatures on
>>the non-official releases?
>>
>> -Mark
>
RE: Convenience Binary Policy
Posted by Kessler CTR Mark J <ma...@usmc.mil>.
The only reason I bring this up; some people verify the package has not been modified if getting it from another source.
-Mark
-----Original Message-----
From: Harbs [mailto:harbs.lists@gmail.com]
Sent: Thursday, October 23, 2014 10:00 AM
To: dev@flex.apache.org
Subject: Re: Convenience Binary Policy
I think that's up to us.
On Oct 23, 2014, at 4:56 PM, Kessler CTR Mark J <ma...@usmc.mil> wrote:
> I like the idea so far. Would there still be hashes / signatures on the non-official releases?
>
> -Mark
Re: Convenience Binary Policy
Posted by Harbs <ha...@gmail.com>.
I think that’s up to us.
On Oct 23, 2014, at 4:56 PM, Kessler CTR Mark J <ma...@usmc.mil> wrote:
> I like the idea so far. Would there still be hashes / signatures on the non-official releases?
>
> -Mark
RE: Convenience Binary Policy
Posted by Kessler CTR Mark J <ma...@usmc.mil>.
I like the idea so far. Would there still be hashes / signatures on the non-official releases?
-Mark
-----Original Message-----
From: Harbs [mailto:harbs.lists@gmail.com]
Sent: Thursday, October 23, 2014 9:48 AM
To: dev@flex.apache.org
Subject: Re: Convenience Binary Policy
Just a bit of background here for anyone not following the discussion on @general[1]
This was prompted by the issue with the failing FlexJS installer. The general consensus at Apache is that releases are for source code only. "convenience packages may be provided as well, but they are not official releases.
However, when these convenience packages are distributed by Apache, the waters become muddled. Since it becomes unclear whether the convenience packages are under the umbrella of the Apache release, it's not clear whether changes to them require votes and what official process needs to be followed.
The solution that was proposed on @general was to make a clear distinction between the official release of the source and any convenience packages. To make the distinction clear, the idea was that it would not be hosted by Apache at all and we'd be clear that any installers (that would make everyone's life easier) is not an official Apache release. It seems to me that Github is the logical place to host this effort.
The advantages I see with this is that the release process should become simpler. As soon as the Source passes muster, we could make the release official, and installers would not be official releases at all. We could fix/modify installers and the like as much as needed without worrying about legal implications.
It would also give us a place to host any non-Apache compatible "extras" (such as flexmojo).
[1]http://mail-archives.apache.org/mod_mbox/incubator-general/201410.mbox/%3CD06AD5EC.1BD2C%25aharui@adobe.com%3E
On Oct 23, 2014, at 4:32 PM, Harbs <ha...@gmail.com> wrote:
> Sure. Let's discuss away! :-) (Nothing has changed yet.)
>
> The repo *is* public, and my intention was that we'd give all Flex committers write access.
>
> On Oct 23, 2014, at 4:28 PM, Justin Mclean <ju...@classsoftware.com> wrote:
>
>> Hi,
>>
>> Moving to flex dev list from general.
>>
>>> I just created "flex-extras" and we'll put in some appropriate text both on the Github side and Apache Flex side. Hopefully this will prove to be a smooth resolution.
>>
>>
>> I know you mean well but this really needs to be discussed on the dev list/with the PMC before making any changes.
>>
>> If we were to move to this model it also would be best if the repo was public.
>>
>> Thanks,
>> Justin
>
Re: Convenience Binary Policy
Posted by Harbs <ha...@gmail.com>.
Just a bit of background here for anyone not following the discussion on @general[1]
This was prompted by the issue with the failing FlexJS installer. The general consensus at Apache is that releases are for source code only. “convenience packages may be provided as well, but they are not official releases.
However, when these convenience packages are distributed by Apache, the waters become muddled. Since it becomes unclear whether the convenience packages are under the umbrella of the Apache release, it’s not clear whether changes to them require votes and what official process needs to be followed.
The solution that was proposed on @general was to make a clear distinction between the official release of the source and any convenience packages. To make the distinction clear, the idea was that it would not be hosted by Apache at all and we’d be clear that any installers (that would make everyone’s life easier) is not an official Apache release. It seems to me that Github is the logical place to host this effort.
The advantages I see with this is that the release process should become simpler. As soon as the Source passes muster, we could make the release official, and installers would not be official releases at all. We could fix/modify installers and the like as much as needed without worrying about legal implications.
It would also give us a place to host any non-Apache compatible “extras” (such as flexmojo).
[1]http://mail-archives.apache.org/mod_mbox/incubator-general/201410.mbox/%3CD06AD5EC.1BD2C%25aharui@adobe.com%3E
On Oct 23, 2014, at 4:32 PM, Harbs <ha...@gmail.com> wrote:
> Sure. Let’s discuss away! :-) (Nothing has changed yet.)
>
> The repo *is* public, and my intention was that we’d give all Flex committers write access.
>
> On Oct 23, 2014, at 4:28 PM, Justin Mclean <ju...@classsoftware.com> wrote:
>
>> Hi,
>>
>> Moving to flex dev list from general.
>>
>>> I just created “flex-extras” and we’ll put in some appropriate text both on the Github side and Apache Flex side. Hopefully this will prove to be a smooth resolution.
>>
>>
>> I know you mean well but this really needs to be discussed on the dev list/with the PMC before making any changes.
>>
>> If we were to move to this model it also would be best if the repo was public.
>>
>> Thanks,
>> Justin
>
Re: Convenience Binary Policy
Posted by Harbs <ha...@gmail.com>.
That’s right. I just created the organization. I did not yet create any repos. I wanted input from everyone else on how we’d structure it before we did that.
On Oct 23, 2014, at 4:47 PM, Justin Mclean <ju...@classsoftware.com> wrote:
> Currently I'm seeing:
> "his organization has no public repositories."
Re: Convenience Binary Policy
Posted by Harbs <ha...@gmail.com>.
On Oct 24, 2014, at 10:49 AM, Justin Mclean <ju...@classsoftware.com> wrote:
> Hi,
>
>> It was not Bertrand that said that. It was Brane.
>
> Sorry my mistake, it been a long thread.
No problem. Glad to clear that up.
>
>> There are no issues with having an EXTERNAL dependency any more than there’s a concern of having a dependency on an operating system.
>
> Unless it's Category X [1], that's clear at least. Although there are some exceptions there re build and test tools.
My reading is that even category x is okay as long as it’s not distributed by Apache, but that’s an academic point right now. If we ever have to deal with category x, we can discuss it. (There’s also the subject of different classifications of dependencies: dependencies compiled into the binary, compile time dependencies not compiled in and runtime dependencies. But this is all academic right now...)
Harbs
Re: Convenience Binary Policy
Posted by Justin Mclean <ju...@classsoftware.com>.
Hi,
> It was not Bertrand that said that. It was Brane.
Sorry my mistake, it been a long thread.
> There are no issues with having an EXTERNAL dependency any more than there’s a concern of having a dependency on an operating system.
Unless it's Category X [1], that's clear at least. Although there are some exceptions there re build and test tools.
Thanks,
Justin
1. http://www.apache.org/legal/resolved.html#category-x
2. http://www.apache.org/legal/resolved.html#prohibited
Re: Convenience Binary Policy
Posted by Harbs <ha...@gmail.com>.
It was not Bertrand that said that. It was Brane.
It was a question which I answered by giving a background on JBurg. Case closed.
I understood the question as Alex did that he was confused on how we could be compiling Cat B in our distributed binaries (which is not the case).
There are no issues with having an EXTERNAL dependency any more than there’s a concern of having a dependency on an operating system.
Let’s please not make issues where none exist.
Harbs
On Oct 24, 2014, at 10:19 AM, Justin Mclean <ju...@classsoftware.com> wrote:
> Hi,
>
>> We can go ask the poster what he meant, but I am going to assume he is either mistaken
>
> Bertrand (an ex mentor of our project) usually knows what he talking about. I have seen this optional ve mandatory mentioned a few times elsewhere as well, but was unable to find a clear definition / example in the mailing list archives.
>
> I know what a Category B licences is, a simple link to the Apache docs would of sufficed for other people rather than copying the whole text. It's nice to try and keep emails short.
>
>> Also, note that the main SDK has a required Category B binary dependency on the OSMF.swc.
>
> While we've stated osmf as required leaving it out just breaks the spark video player so it could possibly be considered optional. I don't know what the answer is here or understand the mandatory/optional distinction which is why I asked.
>
> Also on email etiquette it would be good if people refrain from formatting emails with very short line lengths it makes them hard to read. (In this case your reply didn't fit on a single screen). All modern email clients support line wrapping / resizable windows and I think we're past the days of reading email in a 80x25 terminal window :-)
>
> Thanks,
> Justin
Re: Convenience Binary Policy
Posted by Alex Harui <ah...@adobe.com>.
On 10/24/14, 12:19 AM, "Justin Mclean" <ju...@classsoftware.com> wrote:
>Hi,
>
>> We can go ask the poster what he meant, but I am going to assume he is
>>either mistaken
>
>Bertrand (an ex mentor of our project) usually knows what he talking
>about.
My records show it was not Bertrand.
>
>While we've stated osmf as required leaving it out just breaks the spark
>video player so it could possibly be considered optional. I don't know
>what the answer is here or understand the mandatory/optional distinction
>which is why I asked.
Technically, it is my understanding that the current build scripts in the
source package require the OSMF.swc because the Video components that
reference OSMF are in the Spark.swc. but I agree that the binary package
does not require it.
The inclusion of OSMF was reviewed and approved by the mentors including
Bertrand, so I’m confident we are handling this correctly.
-Alex
Re: Convenience Binary Policy
Posted by Justin Mclean <ju...@classsoftware.com>.
Hi,
> We can go ask the poster what he meant, but I am going to assume he is either mistaken
Bertrand (an ex mentor of our project) usually knows what he talking about. I have seen this optional ve mandatory mentioned a few times elsewhere as well, but was unable to find a clear definition / example in the mailing list archives.
I know what a Category B licences is, a simple link to the Apache docs would of sufficed for other people rather than copying the whole text. It's nice to try and keep emails short.
> Also, note that the main SDK has a required Category B binary dependency on the OSMF.swc.
While we've stated osmf as required leaving it out just breaks the spark video player so it could possibly be considered optional. I don't know what the answer is here or understand the mandatory/optional distinction which is why I asked.
Also on email etiquette it would be good if people refrain from formatting emails with very short line lengths it makes them hard to read. (In this case your reply didn't fit on a single screen). All modern email clients support line wrapping / resizable windows and I think we're past the days of reading email in a 80x25 terminal window :-)
Thanks,
Justin
Re: Convenience Binary Policy
Posted by Alex Harui <ah...@adobe.com>.
On 10/23/14, 5:23 PM, "Justin Mclean" <ju...@classsoftware.com> wrote:
>Hi,
>
>> When did he say that?
>
>Here:
>"I have trouble visualising how any ASF project could have /mandatory/
>dependencies on anything from the B-list."
The above quote is from the general@ thread and was not written by anyone
on this list.
>
>I guess it depends on what mandatory means, if it's needed to in order to
>work and isn't optional then we probably have a large issue.
We can go ask the poster what he meant, but I am going to assume he is
either mistaken or is concerned there is a dependency on some source code.
For us, Falcon has a required dependency on a category B binary. I do
not see any information in [1] that says that is not allowed. The
installation of the modified binary package requires that you accept the
CPL license for Jburg before executing the rest of the install. Here’s
the section on Category B, partially condensed:
*How should so-called "Weak Copyleft" Licenses be handled?*
Each license in this category requires some degree of
reciprocity; this may mean that additional action is
warranted in order to minimize the chance that a user of
an Apache product will create a derivative work of a
reciprocally-licensed portion of an Apache product without
being aware of the applicable requirements.
Software under the following licenses may be included
in binary form within an Apache product if the inclusion
is appropriately labeled:
* CDDL, Versions 1.0 and 1.1 <http://www.sun.com/cddl/>
* CPL 1.0 <http://www.opensource.org/licenses/cpl1.0.php>
(I shortened this list)
By including only the object/binary form, there is less
exposed surface area of the third-party work from which
a work might be derived; this addresses the second
guiding principle of this policy. By attaching a prominent
label to the distribution and requiring an explicit action
by the user to get the reciprocally-licensed source,
users are less likely to be unaware of restrictions
significantly different from those of the Apache License.
Please include the URL to the product's homepage in the
prominent label.
(I removed a paragraph here)
This is colloquially known as the Category B list.
Also, note that the main SDK has a required Category B binary dependency
on the OSMF.swc.
-Alex
[1] http://www.apache.org/legal/resolved.html
Re: Convenience Binary Policy
Posted by Justin Mclean <ju...@classsoftware.com>.
Hi,
> When did he say that?
Here:
"I have trouble visualising how any ASF project could have /mandatory/
dependencies on anything from the B-list."
I guess it depends on what mandatory means, if it's needed to in order to work and isn't optional then we probably have a large issue.
Thanks,
Justin
Re: Convenience Binary Policy
Posted by Harbs <ha...@gmail.com>.
When did he say that?
On Oct 23, 2014, at 4:47 PM, Justin Mclean <ju...@classsoftware.com> wrote:
> As Bertrand stated no ASF project can have mandatory dependancies on something with a category B licence. Am I right is assuming it mandatory?
Re: Convenience Binary Policy
Posted by OmPrakash Muppirala <bi...@gmail.com>.
On Thu, Oct 23, 2014 at 3:03 PM, Justin Mclean <ju...@classsoftware.com>
wrote:
> Hi,
>
> > The problem was not with the dependency per se. It was with getting
> JBurg onto the user’s machine.
>
> Which is done by code in the FlexJS release not the installer (it
> installer.xml file). The installer installs other Apache products ie the
> Flex SDK which don't have a dependancy on JBerg and don't have this issue.
>
> > If the installer was not “officially Apache”, we would not have needed
> any of the discussion we had.
>
> Yes it still would of happened as Alex wanted to modify the FlexJS binary
> hosted at Apache without a vote. If the Flex JS Binary was external to
> Apache then probably not but we would still need to be a discussion around
> is it OK to have a blessed installer installing non blessed artefacts. If
> both the Flex JS Binary and Installer was external then sure but then they
> might as well be a fork and not part of Apache at all.
>
> There are still good reason for hosting convenience binaries at Apache:
> - User trust - they are produced at the same time from the source release
> and hosted at Apache
> - Infrastructure - bandwidth, security, mirrors etc
> - Simple to make release as both source and binary are checked in at the
> same time in the same place
> - The official release process encourages people to check what we're
> releasing 3 +1 votes and all that
>
> > This all seems very much in line with the official Apache policy that
> only source is a release anyway. Since Apache doesn’t want responsibility
> for installers and everything related (i.e. binary dependencies etc.), the
> simplest solution is to keep that separation.
>
> At no point has ever said that Apache doesn't want responsibility for
> installers, the installer consists of source code just like any other
> Apache project, so not sure where you are getting that from. All Apache
> project have some means of installing and while that is usually maven or
> ant other projects do have installers as well (eg Open Office)
>
> So if I understand you correctly in order to not vote on stuff we want
> move both the Installer and FlexJS outside of Apache? That's basically a
> fork.
>
>
I don't think this qualifies as a fork, but you can call it that if you
like. Also, we have moved our committer whiteboards to Github in the past.
Thanks,
Om
> Thanks,
> Justin
Re: Convenience Binary Policy
Posted by Justin Mclean <ju...@classsoftware.com>.
Hi,
> The problem was not with the dependency per se. It was with getting JBurg onto the user’s machine.
Which is done by code in the FlexJS release not the installer (it installer.xml file). The installer installs other Apache products ie the Flex SDK which don't have a dependancy on JBerg and don't have this issue.
> If the installer was not “officially Apache”, we would not have needed any of the discussion we had.
Yes it still would of happened as Alex wanted to modify the FlexJS binary hosted at Apache without a vote. If the Flex JS Binary was external to Apache then probably not but we would still need to be a discussion around is it OK to have a blessed installer installing non blessed artefacts. If both the Flex JS Binary and Installer was external then sure but then they might as well be a fork and not part of Apache at all.
There are still good reason for hosting convenience binaries at Apache:
- User trust - they are produced at the same time from the source release and hosted at Apache
- Infrastructure - bandwidth, security, mirrors etc
- Simple to make release as both source and binary are checked in at the same time in the same place
- The official release process encourages people to check what we're releasing 3 +1 votes and all that
> This all seems very much in line with the official Apache policy that only source is a release anyway. Since Apache doesn’t want responsibility for installers and everything related (i.e. binary dependencies etc.), the simplest solution is to keep that separation.
At no point has ever said that Apache doesn't want responsibility for installers, the installer consists of source code just like any other Apache project, so not sure where you are getting that from. All Apache project have some means of installing and while that is usually maven or ant other projects do have installers as well (eg Open Office)
So if I understand you correctly in order to not vote on stuff we want move both the Installer and FlexJS outside of Apache? That's basically a fork.
Thanks,
Justin
Re: Convenience Binary Policy
Posted by Harbs <ha...@gmail.com>.
The problem was not with the dependency per se. It was with getting JBurg onto the user’s machine.
If the installer was not “officially Apache”, we would not have needed any of the discussion we had. We could have simply fixed things on Github and gone on with life. I’m looking to help make things simpler and leave less issues open to complicate the release process. The less bureaucracy required, the better in my book.
This all seems very much in line with the official Apache policy that only source is a release anyway. Since Apache doesn’t want responsibility for installers and everything related (i.e. binary dependencies etc.), the simplest solution is to keep that separation.
Harbs
On Oct 23, 2014, at 4:47 PM, Justin Mclean <ju...@classsoftware.com> wrote:
> So can you explain why you want to move the installer out of Apache? It's FlexJS that has the dependancy on JBerg not the installer. The installer didn't have to be modified to "correct" the FlexJS / JBerg issue.
Re: Convenience Binary Policy
Posted by Justin Mclean <ju...@classsoftware.com>.
Hi,
> Sure. Let’s discuss away! :-) (Nothing has changed yet.)
So can you explain why you want to move the installer out of Apache? It's FlexJS that has the dependancy on JBerg not the installer. The installer didn't have to be modified to "correct" the FlexJS / JBerg issue.
BTW if JBerg is a mandatory dependancy for FlexJS or rather Falcon that means we probably need to remove the Falcon releases and either host it outside Apache or remove the JBerg dependancy from it. As Bertrand stated no ASF project can have mandatory dependancies on something with a category B licence. Am I right is assuming it mandatory?
> The repo *is* public, and my intention was that we’d give all Flex committers write access
Currently I'm seeing:
"his organization has no public repositories."
Thanks,
Justin
Re: Convenience Binary Policy
Posted by Harbs <ha...@gmail.com>.
Sure. Let’s discuss away! :-) (Nothing has changed yet.)
The repo *is* public, and my intention was that we’d give all Flex committers write access.
On Oct 23, 2014, at 4:28 PM, Justin Mclean <ju...@classsoftware.com> wrote:
> Hi,
>
> Moving to flex dev list from general.
>
>> I just created “flex-extras” and we’ll put in some appropriate text both on the Github side and Apache Flex side. Hopefully this will prove to be a smooth resolution.
>
>
> I know you mean well but this really needs to be discussed on the dev list/with the PMC before making any changes.
>
> If we were to move to this model it also would be best if the repo was public.
>
> Thanks,
> Justin
Re: Convenience Binary Policy
Posted by Justin Mclean <ju...@classsoftware.com>.
Hi,
Moving to flex dev list from general.
> I just created “flex-extras” and we’ll put in some appropriate text both on the Github side and Apache Flex side. Hopefully this will prove to be a smooth resolution.
I know you mean well but this really needs to be discussed on the dev list/with the PMC before making any changes.
If we were to move to this model it also would be best if the repo was public.
Thanks,
Justin
Re: Convenience Binary Policy
Posted by Harbs <ha...@gmail.com>.
Great!
I just created “flex-extras” and we’ll put in some appropriate text both on the Github side and Apache Flex side. Hopefully this will prove to be a smooth resolution.
Harbs
On Oct 23, 2014, at 1:22 PM, Bertrand Delacretaz <bd...@apache.org> wrote:
> Hi,
>
> On Thu, Oct 23, 2014 at 11:44 AM, Harbs <ha...@gmail.com> wrote:
>> ...What about a Github repository for “Flex Support” or “Flex Extensions”? We could build installers and
>> host them there without “official blessings”....
>
> That's pretty much what I meant.
>
> As for naming "Foo Extensions for Apache Flex" is better than "Flex
> Foo Extensions" as the former makes it clearer than there's no
> affiliation.
>
>> ...Would it be okay to link to such a repository on the Flex download page with a
>> disclaimer that the installers are not blessed by Apache?...
>
> As long as you state true facts it's fine, many of our projects point
> to outside community resources.
>
> -Bertrand
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
> For additional commands, e-mail: general-help@incubator.apache.org
>
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
For additional commands, e-mail: general-help@incubator.apache.org
Re: Convenience Binary Policy
Posted by Bertrand Delacretaz <bd...@apache.org>.
Hi,
On Thu, Oct 23, 2014 at 11:44 AM, Harbs <ha...@gmail.com> wrote:
> ...What about a Github repository for “Flex Support” or “Flex Extensions”? We could build installers and
> host them there without “official blessings”....
That's pretty much what I meant.
As for naming "Foo Extensions for Apache Flex" is better than "Flex
Foo Extensions" as the former makes it clearer than there's no
affiliation.
> ...Would it be okay to link to such a repository on the Flex download page with a
> disclaimer that the installers are not blessed by Apache?...
As long as you state true facts it's fine, many of our projects point
to outside community resources.
-Bertrand
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
For additional commands, e-mail: general-help@incubator.apache.org
Re: Convenience Binary Policy
Posted by Harbs <ha...@gmail.com>.
Here’s another idea along these lines:
What about a Github repository for “Flex Support” or “Flex Extensions”? We could build installers and host them there without “official blessings”. Any non Apache license compliant extensions (such as flexmojo) might go there as well.
This might just be a workable solution and side-step this whole issue. Would it be okay to link to such a repository on the Flex download page with a disclaimer that the installers are not blessed by Apache?
Harbs
On Oct 23, 2014, at 12:34 PM, Bertrand Delacretaz <bd...@apache.org> wrote:
> Hi,
>
> On Thu, Oct 23, 2014 at 11:14 AM, Harbs <ha...@gmail.com> wrote:
>> JBurg can be installed manually by the end user, but the Flex community has tried to simplify
>> the setup process by downloading JBurg by an installer script...
>
> The easiest might be to host the Flex installer outside of Apache
> Flex, as a separate project where its contributors are free to do
> whatever. This would make it clearer that the installer is not an
> Apache thing, and that it might not offer the same level of confidence
> that our source releases, not the same clean "nothing worse than
> Apache license inside" guarantee.
>
> We used to have http://community.apache.org/apache-extras/faq.html for
> such things but AFAIK people cannot host downloads there anymore so it
> might not be useful here. But the idea is the same - cleanly separate
> the Apache project from extras that might be useful to some people,
> but don't give the same guarantees.
>
> -Bertrand
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
> For additional commands, e-mail: general-help@incubator.apache.org
>
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
For additional commands, e-mail: general-help@incubator.apache.org
Re: Convenience Binary Policy
Posted by Bertrand Delacretaz <bd...@apache.org>.
Hi,
On Thu, Oct 23, 2014 at 11:14 AM, Harbs <ha...@gmail.com> wrote:
> JBurg can be installed manually by the end user, but the Flex community has tried to simplify
> the setup process by downloading JBurg by an installer script...
The easiest might be to host the Flex installer outside of Apache
Flex, as a separate project where its contributors are free to do
whatever. This would make it clearer that the installer is not an
Apache thing, and that it might not offer the same level of confidence
that our source releases, not the same clean "nothing worse than
Apache license inside" guarantee.
We used to have http://community.apache.org/apache-extras/faq.html for
such things but AFAIK people cannot host downloads there anymore so it
might not be useful here. But the idea is the same - cleanly separate
the Apache project from extras that might be useful to some people,
but don't give the same guarantees.
-Bertrand
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
For additional commands, e-mail: general-help@incubator.apache.org
Re: Convenience Binary Policy
Posted by Harbs <ha...@gmail.com>.
Our situation is with JBurg.[1]
We have an ActionScript complier (called Falcon) which was donated along with Flex by Adobe. The compilation has a dependency on JBurg to walk the tree.
One flavor of Flex (called FlexJS) is targeting Javascript and cross-compiling ActionScript code to Javascript code using the Javascript flavor of the Falcon compiler to do so on the end user’s machine. This compilation has a dependency on JBurg. JBurg can be installed manually by the end user, but the Flex community has tried to simplify the setup process by downloading JBurg by an installer script. The script was failing because of a hiccup with getting JBurg from Sourceforge. We wanted to resolve the issue by including the JBurg binary directly.
That’s where all the confusion started… ;-)
Harbs
[1]http://jburg.sourceforge.net/
On Oct 23, 2014, at 11:46 AM, brane@apache.org wrote:
> On 22.10.2014 03:02, Justin Mclean wrote:
>> Hi,
>>
>>> Binary dependencies are, by definition, not released by the ASF; because
>>> we release source code. Also, software that has dependencies that are
>>> only available in binary form is not open-source, in my book.
>> You may possibly be forgetting about Category B licensed dependancies. These may only be included in binary form in an Apache product. [1] They would be still be consider open source by most people :-) The source does exist it's just not able to be included in a Apache product due to the weak copy left provision the licences include.
>
> I have trouble visualising how any ASF project could have /mandatory/
> dependencies on anything from the B-list.
>
> -- Brane
>
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
For additional commands, e-mail: general-help@incubator.apache.org
Re: Convenience Binary Policy
Posted by br...@apache.org.
On 22.10.2014 03:02, Justin Mclean wrote:
> Hi,
>
>> Binary dependencies are, by definition, not released by the ASF; because
>> we release source code. Also, software that has dependencies that are
>> only available in binary form is not open-source, in my book.
> You may possibly be forgetting about Category B licensed dependancies. These may only be included in binary form in an Apache product. [1] They would be still be consider open source by most people :-) The source does exist it's just not able to be included in a Apache product due to the weak copy left provision the licences include.
I have trouble visualising how any ASF project could have /mandatory/
dependencies on anything from the B-list.
-- Brane
Re: Convenience Binary Policy
Posted by Justin Mclean <ju...@classsoftware.com>.
Hi,
> Binary dependencies are, by definition, not released by the ASF; because
> we release source code. Also, software that has dependencies that are
> only available in binary form is not open-source, in my book.
You may possibly be forgetting about Category B licensed dependancies. These may only be included in binary form in an Apache product. [1] They would be still be consider open source by most people :-) The source does exist it's just not able to be included in a Apache product due to the weak copy left provision the licences include.
Thanks,
Justin
1. http://www.apache.org/legal/resolved.html#category-b
Re: Convenience Binary Policy
Posted by Branko Čibej <br...@apache.org>.
On 21.10.2014 15:55, Harbs wrote:
> The one thing I see missing from the proposed text is dependencies and installers.
>
> Particularly this section:
> ### Compiled packages ### {#compiled-packages}
>
> The Apache Software Foundation produces open source software. All releases
> are in the form of the source materials needed to make changes to the
> software being released.
>
> As a convenience to users that might not have the appropriate tools to build a
> compiled version of the source, binary/bytecode packages MAY be distributed
> alongside official Apache releases. In all such cases, the
> binary/bytecode package MUST have the same version number as the source
> release and MUST only add binary/bytecode files that are the result of
> compiling that version of the source code release.
> ---
> How do binary dependencies fit in?
Binary dependencies are, by definition, not released by the ASF; because
we release source code. Also, software that has dependencies that are
only available in binary form is not open-source, in my book.
-- Brane
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
For additional commands, e-mail: general-help@incubator.apache.org
Re: Convenience Binary Policy
Posted by Harbs <ha...@gmail.com>.
The one thing I see missing from the proposed text is dependencies and installers.
Particularly this section:
### Compiled packages ### {#compiled-packages}
The Apache Software Foundation produces open source software. All releases
are in the form of the source materials needed to make changes to the
software being released.
As a convenience to users that might not have the appropriate tools to build a
compiled version of the source, binary/bytecode packages MAY be distributed
alongside official Apache releases. In all such cases, the
binary/bytecode package MUST have the same version number as the source
release and MUST only add binary/bytecode files that are the result of
compiling that version of the source code release.
---
How do binary dependencies fit in?
On Oct 21, 2014, at 4:43 PM, Alex Harui <ah...@adobe.com> wrote:
>
>
> On 10/21/14, 5:57 AM, "Marvin Humphrey" <ma...@rectangular.com> wrote:
>
>>
>> The problem is that we lack a concise policy document. That's where the
>> "ASF
>> release policy codification proposal" as worked through on legal-discuss
>> a few
>> months ago is supposed to help.
>>
>> http://s.apache.org/aGm
>> https://github.com/rectang/asfrelease
>
> Hi Marvin, IMO, this version of the document also doesn’t prohibit me from
> updating the binary package without a vote. It seems like some portions
> of this document is supposed to apply to binary packages like the
> LICENSE/NOTICE sections so I think that’s why there is still ongoing
> debate on the Flex mailing list about whether I can in fact modify the
> FlexJS binary package.
>
> At this point, unless someone not on the Flex PMC says that we can’t make
> the modifications, I am planning to add a new entry to the Flex Installer
> that points to bits that are hosted on my personal server called “Apache
> FlexJS 0.0.2 (with Jburg)” and expose it to the general public. All of
> the compiled bits are the same, but I have added the jburg.jar to the
> package to workaround the problem. The original binary package will
> remain on dist/mirrors.
>
> Thanks,
> -Alex
>
> B�KKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKCB��[��X��ܚX�KK[XZ[��[�\�[][��X��ܚX�P[��X�]܋�\X�K�ܙ�B��܈Y][ۘ[��[X[��K[XZ[��[�\�[Z[[��X�]܋�\X�K�ܙ�B
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
For additional commands, e-mail: general-help@incubator.apache.org
Re: Convenience Binary Policy
Posted by Marvin Humphrey <ma...@rectangular.com>.
On Tue, Oct 21, 2014 at 6:43 AM, Alex Harui <ah...@adobe.com> wrote:
> At this point, unless someone not on the Flex PMC says that we can’t make
> the modifications,
I understand that your deadline is today, thus you are operating in emergency
mode and within what you have determined individually is within the bounds of
your authority. Very well.
However, while opinions and suggestions from outsiders will hopefully prove
helpful, it's the Flex PMC which has been charged with stewardship of Flex,
not the subscribership to general@incubator. It's up to the Flex PMC to
decide whether to continue pursuing the matter.
Marvin Humphrey
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
For additional commands, e-mail: general-help@incubator.apache.org
Re: Convenience Binary Policy
Posted by Alex Harui <ah...@adobe.com>.
On 10/21/14, 5:57 AM, "Marvin Humphrey" <ma...@rectangular.com> wrote:
>
>The problem is that we lack a concise policy document. That's where the
>"ASF
>release policy codification proposal" as worked through on legal-discuss
>a few
>months ago is supposed to help.
>
> http://s.apache.org/aGm
> https://github.com/rectang/asfrelease
Hi Marvin, IMO, this version of the document also doesn’t prohibit me from
updating the binary package without a vote. It seems like some portions
of this document is supposed to apply to binary packages like the
LICENSE/NOTICE sections so I think that’s why there is still ongoing
debate on the Flex mailing list about whether I can in fact modify the
FlexJS binary package.
At this point, unless someone not on the Flex PMC says that we can’t make
the modifications, I am planning to add a new entry to the Flex Installer
that points to bits that are hosted on my personal server called “Apache
FlexJS 0.0.2 (with Jburg)” and expose it to the general public. All of
the compiled bits are the same, but I have added the jburg.jar to the
package to workaround the problem. The original binary package will
remain on dist/mirrors.
Thanks,
-Alex
Re: Convenience Binary Policy
Posted by Harbs <ha...@gmail.com>.
This all sounds very good to me and makes a lot of sense.
As newcomers to Apache, Flex has had a lot of confusion over what “should” and “must” be done regarding releases. If these things would be spelled out more explicitly, it would help current and future Apache projects get it right with minimum disagreement.
Harbs
On Oct 23, 2014, at 10:11 AM, Bertrand Delacretaz <bd...@apache.org> wrote:
> Hi,
>
> On Thursday, October 23, 2014, Roman Shaposhnik <ro...@shaposhnik.org>
> wrote:
>
>> ...I understand the need of projects like OO to provide binaries of some
> sort,
>> I just don't understand why do they have to be 'blessed' by ASF. Once
>> source gets built and packaged a whole new set of issues kick in. I don't
>> think the foundation is well prepared to deal with those. We might as
>> well admit it explicitly...
>
> My understanding is that while we don't make any guarantees about
> convenience binaries, and while they are not ASF releases, it is good to be
> able to verify that the binary that you got is the one that someone from
> the PMC prepared.
>
> So if our PMCs distribute convenience binaries, signing them for example is
> a good thing to allow users to verify that they are using what the PMC
> built, as opposed to some rogue binary. Signing doesn't mean the binary is
> "blessed", it just allows users to verify that they are using what the PMC
> intended to distribute.
>
> -Bertrand
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
For additional commands, e-mail: general-help@incubator.apache.org
Re: Convenience Binary Policy
Posted by Bertrand Delacretaz <bd...@apache.org>.
Hi,
On Thursday, October 23, 2014, Roman Shaposhnik <ro...@shaposhnik.org>
wrote:
> ...I understand the need of projects like OO to provide binaries of some
sort,
> I just don't understand why do they have to be 'blessed' by ASF. Once
> source gets built and packaged a whole new set of issues kick in. I don't
> think the foundation is well prepared to deal with those. We might as
> well admit it explicitly...
My understanding is that while we don't make any guarantees about
convenience binaries, and while they are not ASF releases, it is good to be
able to verify that the binary that you got is the one that someone from
the PMC prepared.
So if our PMCs distribute convenience binaries, signing them for example is
a good thing to allow users to verify that they are using what the PMC
built, as opposed to some rogue binary. Signing doesn't mean the binary is
"blessed", it just allows users to verify that they are using what the PMC
intended to distribute.
-Bertrand
Re: Convenience Binary Policy
Posted by Ted Dunning <te...@gmail.com>.
On Wed, Oct 22, 2014 at 9:37 PM, Roman Shaposhnik <ro...@shaposhnik.org>
wrote:
> I understand the need of projects like OO to provide binaries of some sort,
> I just don't understand why do they have to be 'blessed' by ASF. Once
> source gets built and packaged a whole new set of issues kick in. I don't
> think the foundation is well prepared to deal with those. We might as
> well admit it explicitly.
>
AOO has very special needs for signing that are expressed well later in
another sub-thread.
Another aspect of binary releases is that even if you don't really need a
blessed distro, there are several aspects of the blessing ritual which are
very helpful. For Apache Drill, these include:
1) if I produce a release, it *really* helps to have several eyes beside my
own on the process. Even if I never claim that this is a blessed release
and I never provide my binary release via any Apache channel, it is a huge
benefit to have the community help ensure the quality of the release.
Having a community blessed procedure to build a binary also increases the
probable quality of any commercial versions. Quality here means mostly
license provenance quality, which is something often overlooked by
commercial builds.
2) it is substantially in the project's interest to both (a) have links to
binary convenience packages, and (b) have links to those packages on the
project web site. As a PMC member, I am very uneasy about providing links
to packages that the PMC has not inspected carefully. As such, the
blessing ritual serves an important role here as well.
Different projects will have very different needs, of course, but Drill
definitely finds value in careful inspection of binary artifacts.
RE: Convenience Binary Policy
Posted by "Dennis E. Hamilton" <de...@acm.org>.
<orcnote> below,
-----Original Message-----
From: shaposhnik@gmail.com [mailto:shaposhnik@gmail.com] On Behalf Of Roman Shaposhnik
Sent: Wednesday, October 22, 2014 21:37
To: general@incubator.apache.org
Subject: Re: Convenience Binary Policy
On Tue, Oct 21, 2014 at 5:57 AM, Marvin Humphrey <ma...@rectangular.com> wrote:
> On Mon, Oct 20, 2014 at 10:26 PM, Roman Shaposhnik <ro...@shaposhnik.org> wrote:
>
>>> P.S.: Why anyone would think voting on binaries makes any kind of sense
>>> around here is, of course, a different question. I can't even begin to
>>> count the number of times it's been pointed out that binaries are not
>>> Apache releases.
>>
>> And yet that issue keeps rearing its ugly head. Given the amount
>> of traffic I've seen around clarifying some finer (and not so fine)
>> points of release/licensing implication it is about time we start
>> an FAQ on that. Me thinks at least.
>
> We have several release FAQs spread across the ASF, though -- which, taken
> together, are comprehensive beyond the point of excess.
>
> The problem is that we lack a concise policy document. That's where the "ASF
> release policy codification proposal" as worked through on legal-discuss a few
> months ago is supposed to help.
>
> http://s.apache.org/aGm
> https://github.com/rectang/asfrelease
>
> It's delayed because I got swamped but it seems that the need has not
> diminished.
It would be really nice if you/we could finish this. That said, I
still maintain,
that focusing exclusively on source releases will simplify our life greatly.
IOW, Apache FOO version x.y.z can only designate a source release.
I understand the need of projects like OO to provide binaries of some sort,
I just don't understand why do they have to be 'blessed' by ASF. Once
source gets built and packaged a whole new set of issues kick in. I don't
think the foundation is well prepared to deal with those. We might as
well admit it explicitly.
<orcnote>
That's fine. However the Apache OpenOffice project has an important
need to establish the provenance and integrity of the end-user
binary distributions it makes. There are an incredible variety of
poseur builds that are distributed by third parties for the purpose
of embedding adware. These also pose as "official" releases, including
all of the links to support, pinging for updates, etc. When folks come
with complaints, the project lists refer them to the authentic builds
that the project curates. This is a significant portion of support
requests to users @oo.a.o and dev @oo.a.o.
Now that there is code-signing for the Windows (and Mac?) builds,
this is going to become easier to fend off. Having the binary distros
available from the operating-system "store" applications will provide
further safeguards.
It is true that the package systems of *nix systems is helpful
in terms of authenticity and sourcing of OS-integrated builds and
their updates. There is no such source for 80% of the community of
Apache OpenOffice users. Before, that community was served by Sun,
Oracle, Novell, and IBM. Now, but for the Apache OpenOffice project
and The Document Foundation, there are no prominent, responsible parties
for distributions to the Windows and Mac communities of users.
I do agree that this can be considered a project-specific arrangement
and does not need to be under ASF governance as strictly as for source
distributions, so long as the "more-than-merely-convenient" binary
distributions do not reflect badly on the foundation. Guidance would
still be helpful, since there are many projects and podlings need some
touchstone to encourage good behavior in this area.
</orcnote>
Thanks,
Roman.
P.S. To be super explicit: I'm not saying that binary artifacts should be
banned, rather that a binary artifact compiled by a member of a project
is no different from one compiled by RedHat or Debian and thus requires
no special treatment whatsoever.
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
For additional commands, e-mail: general-help@incubator.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
For additional commands, e-mail: general-help@incubator.apache.org
Re: Convenience Binary Policy
Posted by Roman Shaposhnik <ro...@shaposhnik.org>.
On Tue, Oct 21, 2014 at 5:57 AM, Marvin Humphrey <ma...@rectangular.com> wrote:
> On Mon, Oct 20, 2014 at 10:26 PM, Roman Shaposhnik <ro...@shaposhnik.org> wrote:
>
>>> P.S.: Why anyone would think voting on binaries makes any kind of sense
>>> around here is, of course, a different question. I can't even begin to
>>> count the number of times it's been pointed out that binaries are not
>>> Apache releases.
>>
>> And yet that issue keeps rearing its ugly head. Given the amount
>> of traffic I've seen around clarifying some finer (and not so fine)
>> points of release/licensing implication it is about time we start
>> an FAQ on that. Me thinks at least.
>
> We have several release FAQs spread across the ASF, though -- which, taken
> together, are comprehensive beyond the point of excess.
>
> The problem is that we lack a concise policy document. That's where the "ASF
> release policy codification proposal" as worked through on legal-discuss a few
> months ago is supposed to help.
>
> http://s.apache.org/aGm
> https://github.com/rectang/asfrelease
>
> It's delayed because I got swamped but it seems that the need has not
> diminished.
It would be really nice if you/we could finish this. That said, I
still maintain,
that focusing exclusively on source releases will simplify our life greatly.
IOW, Apache FOO version x.y.z can only designate a source release.
I understand the need of projects like OO to provide binaries of some sort,
I just don't understand why do they have to be 'blessed' by ASF. Once
source gets built and packaged a whole new set of issues kick in. I don't
think the foundation is well prepared to deal with those. We might as
well admit it explicitly.
Thanks,
Roman.
P.S. To be super explicit: I'm not saying that binary artifacts should be
banned, rather that a binary artifact compiled by a member of a project
is no different from one compiled by RedHat or Debian and thus requires
no special treatment whatsoever.
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
For additional commands, e-mail: general-help@incubator.apache.org
Re: Convenience Binary Policy
Posted by Marvin Humphrey <ma...@rectangular.com>.
On Mon, Oct 20, 2014 at 10:26 PM, Roman Shaposhnik <ro...@shaposhnik.org> wrote:
>> P.S.: Why anyone would think voting on binaries makes any kind of sense
>> around here is, of course, a different question. I can't even begin to
>> count the number of times it's been pointed out that binaries are not
>> Apache releases.
>
> And yet that issue keeps rearing its ugly head. Given the amount
> of traffic I've seen around clarifying some finer (and not so fine)
> points of release/licensing implication it is about time we start
> an FAQ on that. Me thinks at least.
We have several release FAQs spread across the ASF, though -- which, taken
together, are comprehensive beyond the point of excess.
The problem is that we lack a concise policy document. That's where the "ASF
release policy codification proposal" as worked through on legal-discuss a few
months ago is supposed to help.
http://s.apache.org/aGm
https://github.com/rectang/asfrelease
It's delayed because I got swamped but it seems that the need has not
diminished.
Marvin Humphrey
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
For additional commands, e-mail: general-help@incubator.apache.org
Re: Convenience Binary Policy
Posted by Roman Shaposhnik <ro...@shaposhnik.org>.
On Mon, Oct 20, 2014 at 9:45 PM, Branko Čibej <br...@apache.org> wrote:
> On 21.10.2014 06:34, Alex Harui wrote:
>> What is the piece I’m missing that says we have to vote to update the
>> binary package?
>
> Apparently the Flex community believes that convenience binaries need
> votes. They don't, but aside from that, if you guys are already voting
> on binary packages, it makes perfect sense to vote for your fixed
> version, if only to keep people happy.
>
>
> -- Brane
>
> P.S.: Why anyone would think voting on binaries makes any kind of sense
> around here is, of course, a different question. I can't even begin to
> count the number of times it's been pointed out that binaries are not
> Apache releases.
And yet that issue keeps rearing its ugly head. Given the amount
of traffic I've seen around clarifying some finer (and not so fine)
points of release/licensing implication it is about time we start
an FAQ on that. Me thinks at least.
Thanks,
Roman.
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
For additional commands, e-mail: general-help@incubator.apache.org
Re: Convenience Binary Policy
Posted by Branko Čibej <br...@apache.org>.
On 21.10.2014 06:34, Alex Harui wrote:
> What is the piece I’m missing that says we have to vote to update the
> binary package?
Apparently the Flex community believes that convenience binaries need
votes. They don't, but aside from that, if you guys are already voting
on binary packages, it makes perfect sense to vote for your fixed
version, if only to keep people happy.
-- Brane
P.S.: Why anyone would think voting on binaries makes any kind of sense
around here is, of course, a different question. I can't even begin to
count the number of times it's been pointed out that binaries are not
Apache releases.
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
For additional commands, e-mail: general-help@incubator.apache.org
Re: Convenience Binary Policy
Posted by Ted Dunning <te...@gmail.com>.
On Mon, Oct 20, 2014 at 9:34 PM, Alex Harui <ah...@adobe.com> wrote:
> >Then this is the Acme Software Foundation installer and you can do what
> >you
> >like.
> I suppose we could, but it wouldn’t be easily found by folks who arrive at
> flex.a.o looking for FlexJS. They’ll probably end up using the current
> Installer and getting an error.
I think that the web site could quite reasonably say something like "Here
is a link to a temporary, unofficial installer that may fix a bug". It
should still show the link to the official install as well, of course.
Re: Convenience Binary Policy
Posted by Alex Harui <ah...@adobe.com>.
On 10/20/14, 5:54 PM, "Ted Dunning" <te...@gmail.com> wrote:
>
>Why not just roll your own installer that has these additional options?
>
>Then this is the Acme Software Foundation installer and you can do what
>you
>like.
I suppose we could, but it wouldn’t be easily found by folks who arrive at
flex.a.o looking for FlexJS. They’ll probably end up using the current
Installer and getting an error.
I guess you are saying that there is no quick fix of a convenience binary.
I guess I’ve been reading things like Marvin/Roy in this thread [1] that
says that a binary package isn’t official which made me think we had more
flexibility. Here’s the quote from Marvin quoting Roy:
An official release by the Apache Software Foundation
consists of source code which has been audited by a PMC.
Of course it is not possible to audit an entire codebase
at each release point, but we achieve that effective result
through PMC monitoring of a "commits" list: if the last
release was fully reviewed, each delta since then has also
been reviewed, and we can demonstrate that the difference
between the two releases is the sum of those deltas, then
the current release has been reviewed.
Binaries combine that carefully audited source code with
an opaque build machine, and the result is not audit able.
Releasing source is an "act of the foundation". A binary
package is an act of the individual who prepared it.
The Foundation was not set up to take on the liability
associated with binary releases:
http://s.apache.org/roy-binary-deps-3
How is that different from any of our other projects?
End users don't compile Java. Hell, most developers
don't compile Java. We distribute plenty of binaries.
We just don't call them SOURCE. The source is what we
review. The source is what we bless. If anyone
wants to go further than that, they are free to do so
as long as they don't call the result an Apache release.
It is a binary package, a user convenience, a download
hosted by openoffice.org. I don't care.
And this from Bertrand [2]
To clarify, the ASF only releases source code - votes on
releases are not "more" about source, they are *only*
about source.
What is the piece I’m missing that says we have to vote to update the
binary package?
-Alex
[1]
http://mail-archives.apache.org/mod_mbox/incubator-general/201410.mbox/%3cC
AAS6=7hMpFR-mATBEppEPA0vyBO6N35YFXkdTtnQnMF+6_gFvQ@mail.gmail.com%3e
[2]
http://mail-archives.apache.org/mod_mbox/incubator-general/201410.mbox/%3cC
AEWfVJ=HFp_OC-bYyoKjM0oPm5a_Bxa4+yOzdvfvrs3uFZyJWA@mail.gmail.com%3e
Re: Convenience Binary Policy
Posted by Ted Dunning <te...@gmail.com>.
On Mon, Oct 20, 2014 at 5:21 PM, Alex Harui <ah...@adobe.com> wrote:
> >If he wants to build his own installer, fine. If it says it is
> >downloading
> >an Apache artifact, it should be voted.
> The Installer has a DropDown list of releases, such as “Apache Flex SDK
> 4.13.0” and “Apache FlexJS 0.0.2”. What if the entry that points to the
> modified package says “Apache FlexJS 0.0.2 (unofficial)” or “Unofficial
> Fix for Apache FlexJS 0.0.2”? But it would be the same Installer, not my
> own. Or does that get into the “promoting nightly builds” territory even
> though all the compiled code came from an official release?
>
Why not just roll your own installer that has these additional options?
Then this is the Acme Software Foundation installer and you can do what you
like.
Re: Convenience Binary Policy
Posted by Alex Harui <ah...@adobe.com>.
On 10/20/14, 4:57 PM, "Ted Dunning" <te...@gmail.com> wrote:
>On Mon, Oct 20, 2014 at 4:49 PM, Justin Mclean <ju...@classsoftware.com>
>wrote:
>
>> > 4) you aren't claiming that the artifact you created is an Apache
>>release
>> > and you are pointing some workshop participants at your release.
>>
>> My understanding is Alex does want to use this as an official release
>>and
>> have the officially released Apache Flex installer download and install
>>it.
We don’t have to do that. The bits on dist/mirrors can stay untouched,
and the downloads page won’t change either. The main thing is whether the
Installer app can point somewhere else.
>> The new binary would be available to everyone / the general public and
>>not
>> just the people attending the talk.
This is true. Anyone using the Installer would get the modified binary.
>
>No vote, not official.
>
>If he wants to build his own installer, fine. If it says it is
>downloading
>an Apache artifact, it should be voted.
The Installer has a DropDown list of releases, such as “Apache Flex SDK
4.13.0” and “Apache FlexJS 0.0.2”. What if the entry that points to the
modified package says “Apache FlexJS 0.0.2 (unofficial)” or “Unofficial
Fix for Apache FlexJS 0.0.2”? But it would be the same Installer, not my
own. Or does that get into the “promoting nightly builds” territory even
though all the compiled code came from an official release?
-Alex
Re: Convenience Binary Policy
Posted by Ted Dunning <te...@gmail.com>.
On Mon, Oct 20, 2014 at 4:49 PM, Justin Mclean <ju...@classsoftware.com>
wrote:
> > 4) you aren't claiming that the artifact you created is an Apache release
> > and you are pointing some workshop participants at your release.
>
> My understanding is Alex does want to use this as an official release and
> have the officially released Apache Flex installer download and install it.
> The new binary would be available to everyone / the general public and not
> just the people attending the talk. (I don't think it's a workshop). The
> currently nightly build doesn't have this issue it's only effects the
> previously officially released 0.01 and 0.02 versions of FlexJS which are
> publicly available via the installer and the Apache mirrors.
>
No vote, not official.
If he wants to build his own installer, fine. If it says it is downloading
an Apache artifact, it should be voted.
Re: Convenience Binary Policy
Posted by Justin Mclean <ju...@classsoftware.com>.
Hi,
> 3) you created a correct distribution artifact and put it somewhere
> non-Apache
The modified binary has been placed in his Apache account [1] and AFAIK he wants to move it to the official a.o/dist release area without a vote or alternatively distribute it directly from there (to avoid waiting for the mirrors to catch up). It's currently been added to the installer as a dev only build (for testing purposes). By "correct distribution" it's a modified version of the existing 0.02 release binary. For someone to duplicate this they would have to unpack the existing binary, modify it by adding the Jberg jar and repackage it. No changes to the build process or source code, LICENSE, NOTICE etc have not been committed to the Apache Flex repository.
> 4) you aren't claiming that the artifact you created is an Apache release
> and you are pointing some workshop participants at your release.
My understanding is Alex does want to use this as an official release and have the officially released Apache Flex installer download and install it. The new binary would be available to everyone / the general public and not just the people attending the talk. (I don't think it's a workshop). The currently nightly build doesn't have this issue it's only effects the previously officially released 0.01 and 0.02 versions of FlexJS which are publicly available via the installer and the Apache mirrors.
I have already stated my view on the Flex dev list that we should follow the normal Apache vote and release cycle [3], but rather than repeat that here I think it boils down to just this: [2]
"Under no circumstances are unapproved builds a substitute for releases. If this policy seems inconvenient, then release more often."
The Jberg jar is licensed under CPL so the binary would also would require a change to NOTICE to comply with Apache licensing policy.
Thanks,
Justin
1. http://people.apache.org/~aharui/FlexJS/binaries/0.0.2a/
2. http://www.apache.org/dev/release.html#what
3. http://markmail.org/message/b2wbyo5yzbrb6f7d
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
For additional commands, e-mail: general-help@incubator.apache.org
Re: Convenience Binary Policy
Posted by Ted Dunning <te...@gmail.com>.
On Mon, Oct 20, 2014 at 3:08 PM, Alex Harui <ah...@adobe.com> wrote:
> I know we can’t go messing around with source packages without a vote, but
> what about binary packages? Is it against policy to do something like
> this, and if so, can exceptions be made?
>
I may not have followed this quite correctly, here is what I understood of
the situation as you described it:
1) there is a bug in the FlexJS distro, considered low priority due to
sparse use
2) you needed a quickly corrected binary distribution
3) you created a correct distribution artifact and put it somewhere
non-Apache
4) you aren't claiming that the artifact you created is an Apache release
and you are pointing some workshop participants at your release.
I fail to see any problem whatsoever in what you did. You used Apache
software to create a derived work which you are asking people to use in an
instructional setting. As far as I can tell, the only claim you are making
is that your artifact is FlexJS with a fix that should be incorporated
upstream before long.
What's the problem?