You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@subversion.apache.org by "JS.staff" <js...@ecclescollege.ac.uk> on 2004/10/08 21:26:26 UTC
svn 1.1.1 with r11211 - users perspective
Sorry for posting this to the dev list, but I've read your thread re r11211 with interest.
First let me say that SVN 1.1 is amazing, fast, compact repositories (fsfs), works great. I love it!
I can't understand, however, any reason for delaying a fix for the 'svn ls' slow-down (of TSVN repos browsing). I, for one, use the TSVN repos browser all the time. Hourly.
Many of my project folders take 10-15 *minutes* to list using 1.1!!!!! Many people are using RC1 it's sooo bad. Come on, people prefer to use a RC than the actual release??????? something smells bad...
Guys, there is no loss of face in releasing 1.1.1 now. There is only loss of face in releasing 1.1.0 with such a big bug in it (took me all of 5 minutes to find! I should be an alpha tester! LOL). The embarasment has already happened. Move on.
If you leave it like this for weeks you know your going to get kicked for it. "SVN is soooo sloooow!". When every other aspect of 1.1 is so awsome!!
Shut the moaners up. Release a fix asap. Admit the community screwed up with 1.1.0. Just roll over and pretend in never happened ;)
Thanks for all your hard work, this is meant to be constructive (!),
John
Re: svn 1.1.1 with r11211 - users perspective
Posted by Justin Erenkrantz <ju...@erenkrantz.com>.
--On Tuesday, October 12, 2004 1:02 AM +0200 Tobias Ringström
<to...@ringstrom.mine.nu> wrote:
> Justin Erenkrantz wrote:
>
>> Um, it's hard to vote on something theoretical. If you has produced a
>> release candidate tarball, I'm sure you would have gotten the 3 votes
>> necessary to make 1.1.1 with just that patch in.
>
> Now you lost me. What makes r11211 theoretical, and how would a release
> candidate tarball make more developers vote?
The release was (and still is) theoretical. You suggested a 1.1.1 released
with r11211. There's a difference between talking about an abstract
concept (i.e. desire to have a 1.1.1 release with that fix included) and
then having a concrete release that can be tested (i.e. a file/tarball that
you propose for 1.1.1).
I'll note that this is different than having r11211 merged back to the
1.1.x branch. You still needed to get approval for r11211 to enter 1.1.x.
Yet, once there, you could have produced 1.1.1 candidate tarballs like I'm
suggesting... *shrug* -- justin
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: svn 1.1.1 with r11211 - users perspective
Posted by Tobias Ringström <to...@ringstrom.mine.nu>.
Justin Erenkrantz wrote:
> Um, it's hard to vote on something theoretical. If you has produced a
> release candidate tarball, I'm sure you would have gotten the 3 votes
> necessary to make 1.1.1 with just that patch in.
Now you lost me. What makes r11211 theoretical, and how would a release
candidate tarball make more developers vote?
/Tobias
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: svn 1.1.1 with r11211 - users perspective
Posted by Justin Erenkrantz <ju...@erenkrantz.com>.
--On Monday, October 11, 2004 11:58 PM +0200 Tobias Ringström
<to...@ringstrom.mine.nu> wrote:
> There are two reasons why I did not "take the lead and produce a
> release", the first being that I have only one vote. I thought my email
> would make people start voting, but it didn't work. The second reason is
Um, it's hard to vote on something theoretical. If you has produced a
release candidate tarball, I'm sure you would have gotten the 3 votes
necessary to make 1.1.1 with just that patch in.
> much harder to describe. I really honestly did not think I had the
> mandate to initiate a release without the consent from the long time
> developers. From where I come it's common to get wide acceptance in the
> group before proceeding, so when I heard nothing from people I expected
> to have an opinon I did not know what to do.
Nah. Anyone (even a non-committer) can produce a tarball of Subversion -
that's why we have the dist.sh script. However, only the committers *as a
group* can produce a release. I can't automatically produce a release by
myself. Neither can you. Neither can Ben. But, working together, we can
indeed produce a release. =)
> I'm learning, though. At least I try to... and I appreciate your
> response. I can't learn from silence.
*nod* -- justin
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: svn 1.1.1 with r11211 - users perspective
Posted by kf...@collab.net.
Justin Erenkrantz <ju...@erenkrantz.com> writes:
> I'll comment that if we didn't have a 'trusted' RM who is the same for
> every release, we should be voting to approve a release before it goes
> public. httpd does this (obviously). I think we've let things slide
> here because we have only one RM. But, if more people take the RM
> responsibility (which I think is ultimately a good thing), then a
> release should undergo some sanity tests before we make it live and
> announce it to the public. -- justin
+1
(We can make the appropriate changes to HACKING if/when we cross that
bridge.)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: svn 1.1.1 with r11211 - users perspective
Posted by Justin Erenkrantz <ju...@erenkrantz.com>.
--On Monday, October 11, 2004 6:42 PM -0500 kfogel@collab.net wrote:
> Voting is a very rare last resort. We use voting for two things:
>
> 1. For resolving disputes when we could not consense.
> 2. For choosing which changes to put in an upcoming release.
I'll comment that if we didn't have a 'trusted' RM who is the same for every
release, we should be voting to approve a release before it goes public.
httpd does this (obviously). I think we've let things slide here because we
have only one RM. But, if more people take the RM responsibility (which I
think is ultimately a good thing), then a release should undergo some sanity
tests before we make it live and announce it to the public. -- justin
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: svn 1.1.1 with r11211 - users perspective
Posted by kf...@collab.net.
Tobias Ringström <to...@ringstrom.mine.nu> writes:
> There are two reasons why I did not "take the lead and produce a
> release", the first being that I have only one vote. I thought my
> email would make people start voting, but it didn't work. The second
> reason is much harder to describe. I really honestly did not think I
> had the mandate to initiate a release without the consent from the
> long time developers. From where I come it's common to get wide
> acceptance in the group before proceeding, so when I heard nothing
> from people I expected to have an opinon I did not know what to do.
Voting is a very rare last resort. We use voting for two things:
1. For resolving disputes when we could not consense.
2. For choosing which changes to put in an upcoming release.
(1) hardly ever happens, like, maybe two or three times in the entire
history of this project. (2) is a normal part of the release process,
of course, but it's irrelevant here, since your question is "How do we
decide *whether* to do a release?" not "What should go into a release
once we've decided to do it?"
So, the first thing you should do is: forget about voting. That's not
how we do things, if we can avoid it :-).
On to the real issue: How do we decide whether to make 1.1.1/1.0.9
releases?
Answer: you say
"This regression is so important that I am going to make these
releases unless a full committer vetoes it. Any help is welcome.
I'm willing to discuss alternatives to this strategy, as long as
such discussions are constructive and not mere delaying tactics.
But one way or another, we've got to get a fix out for this bug
ASAP."
Now, can you imagine someone vetoing that :-)? Maybe some people
would register objections, or ask that you go a bit more slowly or
something. Such reactions should be discussed of course, but the
point is to start out from a stance of "This is important to me, and I
am willing to do what it takes to make it happen." That way, you get
the momentum on your side. Most people will get out of your way, or
even help you, once they see a clear intent to act. Don't worry that
you are somehow usurping a right reserved to long-time developers. If
someone has a real problem with it, they'll use their veto. (But I
highly doubt that would happen, since you would be volunteering to do
the work, and we're not talking about a permanent, controversial code
change or something.)
Contrast the above strategy with this one:
"I really think this bug is important enough to warrant a new
release. Does everyone agree with me? Let's make a release."
That's a good way to have a nice conversation, but not a good way to
start momentum toward *actually making* a new release.
It's also a caricature of what you wrote, of course. Please forgive
my oversimplification. I'm just trying to make the point as clear as
possible.
Don't think of yourself as somehow "junior" in this project. You've
been fixing a helluva lot more bugs than I have lately, so please
assume that you have as much right to initiate a release process as I
do. Or as Ben Reser does. Or Brane. Or Greg Hudson. Or... you get
the idea :-).
Best,
-Karl
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: svn 1.1.1 with r11211 - users perspective
Posted by Tobias Ringström <to...@ringstrom.mine.nu>.
Justin Erenkrantz wrote:
> In all honesty, why haven't you (Tobias) taken the lead and produced a
> release that meets your characteristics? At the very least, you could
> have produced a 1.0.9-rc and 1.1.1-rc for testing on this list. There
> should be nothing sacrosanct about Ben Reser doing releases. Having
> only one person who can do releases leads to this constipation.
> That's why dist.sh is so crucial. *shrug* -- justin
Let me first clarify that I absolutely do not mind doing the work. That
has nothing to do with it. Making a release would be much less work than
it was to find the problem and fix it. I was not waiting for Ben to do
the work, and I had absolutely no idea it sounded that way.
There are two reasons why I did not "take the lead and produce a
release", the first being that I have only one vote. I thought my email
would make people start voting, but it didn't work. The second reason is
much harder to describe. I really honestly did not think I had the
mandate to initiate a release without the consent from the long time
developers. From where I come it's common to get wide acceptance in the
group before proceeding, so when I heard nothing from people I expected
to have an opinon I did not know what to do.
I'm learning, though. At least I try to... and I appreciate your
response. I can't learn from silence.
/Tobias
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: svn 1.1.1 with r11211 - users perspective
Posted by Justin Erenkrantz <ju...@erenkrantz.com>.
--On Sunday, October 10, 2004 10:39 PM +0200 Tobias Ringström
<to...@ringstrom.mine.nu> wrote:
> is perfectly alright (although I don't agree, obviously). I think the
> main reason that we've made zero progress towards new releases is that
> the "big guys" (except brane) haven't made any comments (publicly)
> neither for nor against a release. Maybe I should have called a vote.
In all honesty, why haven't you (Tobias) taken the lead and produced a
release that meets your characteristics? At the very least, you could have
produced a 1.0.9-rc and 1.1.1-rc for testing on this list. There should be
nothing sacrosanct about Ben Reser doing releases. Having only one person
who can do releases leads to this constipation. That's why dist.sh is so
crucial. *shrug* -- justin
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: svn 1.1.1 with r11211 - users perspective
Posted by kf...@collab.net.
Tobias Ringström <to...@ringstrom.mine.nu> writes:
> >No vote would have been necessary. Just say "Wow, this problem is
> >really bad, we need a new release, and I volunteer to do whatever's
> >necessary to make that happen, up to and including RM if Ben's busy."
>
> What I did was to say that 1.0.8 and 1.1.0 had a really bad regression
> and that we needed to make new releases ASAP. I asked people to go to
> STATUS and vote.
Well, we use STATUS to vote on the changes that should go in in a
release, not on the question of whether to do a release at all... So
people probably didn't understand what you wanted them to do in
STATUS. (I know I didn't.)
> The only developers who said anything reasoned
> against making a 1.0.x release at all, and against making a quick
> 1.1.x release (which was basically my whole proposal). I will
> certainly not go forward if every other developer who says anything is
> against the idea. I don't see how this has anything to do with Ben as
> the RM, or me offering to be RM for those releases.
Ah. Let me try to explain it better:
The point of offering to do the work (be RM or whatever) is not that
doing so meets some formal guideline. The point is that it puts the
momentum on your side. Think about it psychologically. There are two
ways you can ask the question:
1. "Does everyone agree that we need a release?"
- versus -
2. "Will anyone *stop* me from making a release?"
Technically (1) and (2) are the same question, yes. But asking (2) is
far more far more likely to result in a release, I think you can see
why.
-Karl
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: svn 1.1.1 with r11211 - users perspective
Posted by Tobias Ringström <to...@ringstrom.mine.nu>.
kfogel@collab.net wrote:
>No vote would have been necessary. Just say "Wow, this problem is
>really bad, we need a new release, and I volunteer to do whatever's
>necessary to make that happen, up to and including RM if Ben's busy."
>
>
What I did was to say that 1.0.8 and 1.1.0 had a really bad regression
and that we needed to make new releases ASAP. I asked people to go to
STATUS and vote. The only developers who said anything reasoned against
making a 1.0.x release at all, and against making a quick 1.1.x release
(which was basically my whole proposal). I will certainly not go forward
if every other developer who says anything is against the idea. I don't
see how this has anything to do with Ben as the RM, or me offering to be
RM for those releases.
/Tobias
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: svn 1.1.1 with r11211 - users perspective
Posted by kf...@collab.net.
Tobias Ringström <to...@ringstrom.mine.nu> writes:
> That's great, but don't feel that it's your fault that we don't have
> 1.0.9 or 1.1.1 yet. You spoke against making new releases quickly
> which is perfectly alright (although I don't agree, obviously). I
> think the main reason that we've made zero progress towards new
> releases is that the "big guys" (except brane) haven't made any
> comments (publicly) neither for nor against a release. Maybe I should
> have called a vote.
Aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaargh :-).
If we are to have a "big guys" category in this project, please note
that both you and Ben Reser are as "big" as anyone here. But the
whole concept is slippery, because part of the way someone gets to be
seen as "big" is by taking initiative on things like making releases
happen.
No vote would have been necessary. Just say "Wow, this problem is
really bad, we need a new release, and I volunteer to do whatever's
necessary to make that happen, up to and including RM if Ben's busy."
(Fortunately, Ben can do it soon anyway.)
I don't think anyone would have objected -- partly because when you
volunteer for all that work, it's a sign of how important you think
the issue is :-).
Since Ben is willing to manage it, and only asks for time to get a
handle on the other fixes that might go into the release, it looks
like this has resolved itself. But there's something you can do in
the meantime: Put a notice on our front page describing the ls issue,
reaffirming that it is known & solved, that new releases will be
coming out soon with fixes, and that one can upgrade to trunk head or
apply <link-to-patch/> if one is desparate to fix it sooner.
(I think for a problem this noticeable, a temporary notice like that
is appropriate. Anything else you can think of?)
-Karl, a "big guy" who simply had other things on his plate, and never
meant to hold up a release.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: svn 1.1.1 with r11211 - users perspective
Posted by Tobias Ringström <to...@ringstrom.mine.nu>.
Ben Reser wrote:
>a) Time on my part. I should have time to do 1.0.9 and 1.1.1 this
>coming week.
>
>
That's great, but don't feel that it's your fault that we don't have
1.0.9 or 1.1.1 yet. You spoke against making new releases quickly which
is perfectly alright (although I don't agree, obviously). I think the
main reason that we've made zero progress towards new releases is that
the "big guys" (except brane) haven't made any comments (publicly)
neither for nor against a release. Maybe I should have called a vote.
>b) While the performance issue is annoying it is not a showstopper.
>
>
You make it sound like a fact, but frankly, that's your opinion. It is
most certainly a showstopper for people using "svn ls" or RA->get_dir
(as has been demonstrated several times).
> <>d) People that are finding this issue annoying have several options open
> to them. If they do not need path based authentication with DAV they
> can run 1.1.0-rc3. Practically the only difference between rc3 and rc4
> is the security fix (which causes this problem). And there is only a
> single syntax error fix in 1.1.0 final that was also introduced by the
> security fix. People that do need the security fix can apply the patch.
There are several problem with these options. First of all, they will
have to hit the bug first which is bad in itself, and how are they
supposed to find out which release (candidate) to use? Telling people to
apply patches is also hard, because most of our users are using binaries
which makes applying patches very hard. I'm also certain that we're only
seeing the tip of the iceberg in IRC and users@. There are probably many
people out there muttering and swearing by themselves or in other
forums. If we make a release with a regression as bad as this one
(thinking primarily of 1.0.8 here) without fixing it promptly, we'll
only upset people and make them less keen to upgrade in the future
(which is a problem for us because we don't want bug reports for already
fixed problems, and of course, we want to look good).
In short, our users should not have to pay for our mistakes needlessly.
>I don't forsee any issues with cutting 1.0.9 and 1.1.1 next week
>sometime. But I really do need to get through my email to see if there
>are any outstanding issues.
>
>
Thanks!
/Tobias
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: svn 1.1.1 with r11211 - users perspective
Posted by Ben Reser <be...@reser.org>.
On Sat, Oct 09, 2004 at 11:04:35PM -0500, kfogel@collab.net wrote:
> "JS.staff" <js...@ecclescollege.ac.uk> writes:
> > Guys, there is no loss of face in releasing 1.1.1 now. There is only
> > loss of face in releasing 1.1.0 with such a big bug in it (took me all
> > of 5 minutes to find! I should be an alpha tester! LOL). The
> > embarasment has already happened. Move on.
>
> Thanks for your mail -- it's useful feedback.
>
> I think the real issue is the effort of making another release. Also
> remember that there is a freeze period required for any release --
> only a showstopper bug would cause us to release a ".1" *immediately*
> after a ".0", without the testing period. There's also the question
> of whether a 1.0.9 is appropriate.
Actually we don't do freezes for patch releases. Only major and minor
releases. That said, there are four reasons I haven't cut a 1.1.1:
a) Time on my part. I should have time to do 1.0.9 and 1.1.1 this
coming week.
b) While the performance issue is annoying it is not a showstopper. It
was brought up at the time of the release. While we underestimated the
amount of annoyance it would cause and believed at the time that there
wasn't a fix, we felt it was not a showstopper. As kfogel points out
the only reason we'd cut a 1.1.1 immediately after a 1.1.0 is for a
showstopper.
c) We could cut a 1.1.1 quickly with just this fix. But we know of
several issues with 1.1.0. While less common some of them are harder to
work around or resolve. I believe most of these have been fixed over
this past week. But I need to go through my email and look at all the
issues and dig through the issue tracker to make sure. I do not want to
cut a 1.1.1 only to listen to a whole new group demand that we put out
1.1.2 because we didn't get their issue.
d) People that are finding this issue annoying have several options open
to them. If they do not need path based authentication with DAV they
can run 1.1.0-rc3. Practically the only difference between rc3 and rc4
is the security fix (which causes this problem). And there is only a
single syntax error fix in 1.1.0 final that was also introduced by the
security fix. People that do need the security fix can apply the patch.
I think it's inappropriate to rush a release for a non-showstopper
issue. We'll get the fix out in a release. But let's take the time to
make it a good release.
> (I don't recall anyone saying that loss of face was a reason not to
> release 1.1.1 right away. Are you thinking of a specific mail you
> saw, or are you just guessing about people's motivations?)
I have to agree with kfogel here. But I'll add that not releasing 1.1.1
doesn't save us any face. Nor do I see why anyone would conclude that.
> It's late here, so instead of launching a new thread proposing that we
> do 1.1.1/1.0.9 (and volunteer to manage the releases, since this
> situation isn't the regular RM's fault), I'm going to sleep on it and
> make sure that's something I really want to propose. So forget you
> ever heard this :-). (Of course, someone else could bring it up
> before tomorrow.)
I don't forsee any issues with cutting 1.0.9 and 1.1.1 next week
sometime. But I really do need to get through my email to see if there
are any outstanding issues.
--
Ben Reser <be...@reser.org>
http://ben.reser.org
"Conscience is the inner voice which warns us somebody may be looking."
- H.L. Mencken
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: svn 1.1.1 with r11211 - users perspective
Posted by kf...@collab.net.
"JS.staff" <js...@ecclescollege.ac.uk> writes:
> Guys, there is no loss of face in releasing 1.1.1 now. There is only
> loss of face in releasing 1.1.0 with such a big bug in it (took me all
> of 5 minutes to find! I should be an alpha tester! LOL). The
> embarasment has already happened. Move on.
Thanks for your mail -- it's useful feedback.
I think the real issue is the effort of making another release. Also
remember that there is a freeze period required for any release --
only a showstopper bug would cause us to release a ".1" *immediately*
after a ".0", without the testing period. There's also the question
of whether a 1.0.9 is appropriate.
(I don't recall anyone saying that loss of face was a reason not to
release 1.1.1 right away. Are you thinking of a specific mail you
saw, or are you just guessing about people's motivations?)
It's late here, so instead of launching a new thread proposing that we
do 1.1.1/1.0.9 (and volunteer to manage the releases, since this
situation isn't the regular RM's fault), I'm going to sleep on it and
make sure that's something I really want to propose. So forget you
ever heard this :-). (Of course, someone else could bring it up
before tomorrow.)
-Karl
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org