You are viewing a plain text version of this content. The canonical link for it is here.
Posted to modproxy-dev@apache.org by Eli Marmor <ma...@elmar.co.il> on 2001/02/06 13:17:01 UTC
Punt the Proxy
Hi list!
Are you aware of the discussion at new-httpd@apache.org to punt the
Proxy from the source tree?
Very strong opinions are expressed there, against mod_proxy as a part
of an HTTP server, against mod_proxy as a module without a maintainer,
etc., etc.
I believe that most of us opposed these opinions. On the other hand,
most of us don't have time to join the main list and receive the
zillion messages, so are not aware of that discussion, and almost
nobody tries to argue with the opinions there.
I welcome subscribers of modproxy-dev to join that list (instruction
at http://www.apache.org/foundation/mailinglists.html#http-dev) and
express our side. To see the opinions I mentioned, please look at the
following threads:
http://marc.theaimsgroup.com/?t=97595977400003&w=2&r=1
http://marc.theaimsgroup.com/?l=apache-new-httpd&m=98140196311131&w=2
The best thing we can contribute to the discussion, is somebody who
will take the responsibility to be the official maintainer of this
module.
But even if you can't do it, you can still:
1. Write about the importance of the module.
2. Emphasize the extra importance of the reverse proxy, and that if it
is included, it will make sense to include the whole proxy.
3. Features (including ease, weight, rewrite, etc.) which are not
supported by squid but only by Apache.
4. your experience with it (e.g. "I'm the webmaster of ...... and we
use mod_proxy under 200,000 machines to cache the traffic of our
120 million users. Axing mod_proxy from the source tree, will not
only cause us a problem, but also lower its netcraft statistics
from 60% to 2%").
5. Try to convince that today, with the new mechanism of filtered I/O
in Apache 2.0, mod_proxy is an integral part of Apache more than
ever. IIRC, one of the subscribers here, told about his experience
with porting the proxy to use the filters API, and it sounded
great.
6. Send patches.
7. Tell them about the great success that Apache 2 is going to have:
After 4 years of development, and without key parts such as proxy
and SSL (now even mod_rewrite is under danger!), I'm convinced that
Apache 2 will double netcraft statistics (from 60% to 120%! ;-)
Now seriously (contrary to 7): Apache 2.0 is really going to be a
great thing; It will be pity if the key parts of it will be missing.
So open your mouth and speak.
--
Eli Marmor
marmor@netmask.it
CTO, Founder
Netmask (El-Mar) Internet Technologies Ltd.
__________________________________________________________
Tel.: +972-9-766-1020 8 Yad-Harutzim St.
Fax.: +972-9-766-1314 P.O.B. 7004
Mobile: +972-50-23-7338 Kfar-Saba 44641, Israel
Re: Punt the Proxy
Posted by Ben Hyde <bh...@pobox.com>.
Greg Stein <gs...@lyra.org> writes in response to ben
> > A slight tunning. We are also here to enable users, or thier agents, work
> > on what they desire.
>
> Careful: that is a comment on what you hope the developers feel; I don't
> necessarily believe it is a requirement to be an httpd developer. But it
> doesn't really matter much because I think we have ended up with a
> self-selected group that *does* believe that :-)
Too much thread drift... but yes I reached that conclusion carefully :-)
> > I'm certain there are people out there that really really desire proxy.
...
> Don't get me wrong. I'm definitely for enabling them.
Right, so back to what's the most effective way to trigger some
traction.
The threat, or less colorfully the natural retirement of module
due to the lack of an interested maintainer is an attractive idea.
> I also just get sick of people crying and whining for a feature, and
> *expecting* the Apache developers to do it for them. That expectation is
> severely misguided and exposes a lack of understanding of how our community
> builds software.
hear hear.
Interesting; the threat to toss some functionallity actually helps to
make this case.
The "love it or (we) leave it" tactics best executed with sufficent
fair warning - we do have a lot of pretty locked in users; so the
trick is to give them a means to step forward.
Do we need to be embaressed to say "this part is vistigal for lack
of interest" - probably not. So I'm cool with either booting it
or shipping it labeled as "we intend to boot real soon now".
- ben
Re: Punt the Proxy
Posted by Greg Stein <gs...@lyra.org>.
On Wed, Feb 07, 2001 at 01:06:29PM -0700, Peter J. Cranstone wrote:
> I agree with Ben. There is a great need to think about the end user. I see
> the biggest problem at the moment as the transition of 1.x to 2.x
My point wasn't to code without thinking about the end user. It was that end
users' needs which don't intersect with the current maintainers have no
voice. They can all scream for Nifty Feature Baz, but if a coder never feels
like working on it, then it just isn't going to happen. Of course, some
coders' desire may be to explicitly listen to those screams and code up Baz
whether or not they like doing, just to get the warm fuzzies of making them
happy; but that doesn't contradict what I'm saying: there has to be somebody
interested in coding.
This is a volunteer effort, with people working on the things that interest
them and please them. In many cases, that interest *does* include pleasing
end users, too. But it doesn't mean to please them at all costs, nor please
them at the expense of the developer's enjoyment.
[ insert standard caveats for the people being paid to develop; it is
entirely possible that they would be instructed to code something that
they aren't too happy doing. ]
>...
> My vote doesn't count on this forum, but I would suggest (IMHO) that you
> keep 1.x going for sometime. If it suddenly drops off the radar and 2.0
> doesn't deliver then as more boxes are shipped either old versions of 1.x
> will be used or if the need is for SSL users will switch to IIS.
Oh, don't worry... you input is definitely heard, and counts as another data
point for each members' decisions.
>...
> From: bhyde@zap.ne.mediaone.net [mailto:bhyde@zap.ne.mediaone.net]On Behalf Of Ben Hyde
> Sent: Wednesday, February 07, 2001 12:43 PM
> To: new-httpd@apache.org
> Subject: Re: Punt the Proxy
>
>...
> Greg Stein <gs...@lyra.org> writes:
> > The number of end users is totally irrelevant. It is all about the people
> > *working* on the code. We are not here to satisfy end users' entire scope
> of
> > needs. We are here to satisfy them where their needs intersect our
> desires.
>
> A slight tunning. We are also here to enable users, or thier agents, work
> on what they desire.
Careful: that is a comment on what you hope the developers feel; I don't
necessarily believe it is a requirement to be an httpd developer. But it
doesn't really matter much because I think we have ended up with a
self-selected group that *does* believe that :-)
> I'm certain there are people out there that really really desire proxy.
> They want it bundled in the server - so the trick is to structure things
> so those people get traction on the problem. i.e. so they can hack, not
> whinge, at it.
Don't get me wrong. I'm definitely for enabling them. And I believe we have
that: it is called the module system and the various APIs. Keeping something
in the core is orthogonal to whether it *can* be built. I'm also totally
cool with adding more hooks and APIs to the core to enable mod_proxy
development. I just wish it would leave the core CVS tree; we just keep
doing this over and over again and get nowhere. Move it out... let it live
or die on its own merits. Not because it was forced to (unwillingly) because
of its presence in the core.
I also just get sick of people crying and whining for a feature, and
*expecting* the Apache developers to do it for them. That expectation is
severely misguided and exposes a lack of understanding of how our community
builds software. If they want it, then they can step up to the plate.
Speaking for myself, if all those people demanding the proxy get completed
want to pay me for it, then I might. But as long as my work in contributed
freely, then I'll be contributing the parts that interest me (however varied
those interests may or may not be).
Cheers,
-g
--
Greg Stein, http://www.lyra.org/
RE: Punt the Proxy
Posted by "Peter J. Cranstone" <Cr...@remotecommunications.com>.
I agree with Ben. There is a great need to think about the end user. I see
the biggest problem at the moment as the transition of 1.x to 2.x
Right now you have 10 million + users of a pretty stable code base. All the
modules work or they've been hacked on to work. What's going to make them
shift to 2.0. Multi threading? Maybe if it gives at least a 20% +
performance improvement. The downside is that modules need to be rewritten
and filtering needs to shown to be valuable for everyone to make the shift.
What no one really knows yet is that with multi threading and filtering all
turned on what happens to the promised performance. Again it has to be
something compelling to make people start using it.
My vote doesn't count on this forum, but I would suggest (IMHO) that you
keep 1.x going for sometime. If it suddenly drops off the radar and 2.0
doesn't deliver then as more boxes are shipped either old versions of 1.x
will be used or if the need is for SSL users will switch to IIS.
Peter
-----Original Message-----
From: bhyde@zap.ne.mediaone.net [mailto:bhyde@zap.ne.mediaone.net]On
Behalf Of Ben Hyde
Sent: Wednesday, February 07, 2001 12:43 PM
To: new-httpd@apache.org
Subject: Re: Punt the Proxy
I found Greg's note most amusing.
Greg Stein <gs...@lyra.org> writes:
> The number of end users is totally irrelevant. It is all about the people
> *working* on the code. We are not here to satisfy end users' entire scope
of
> needs. We are here to satisfy them where their needs intersect our
desires.
A slight tunning. We are also here to enable users, or thier agents, work
on what they desire.
I'm certain there are people out there that really really desire proxy.
They want it bundled in the server - so the trick is to structure things
so those people get traction on the problem. i.e. so they can hack, not
whinge,
at it.
- ben
Re: Punt the Proxy
Posted by Nick Kew <ni...@webthing.com>.
Can I offer an outssiders perspective on this, as a user and module
developer but a newbie to new-httpd?
> I'm certain there are people out there that really really desire proxy.
[ steps forward ]
> They want it bundled in the server
Frankly it matters not to me where the proxy is - so long as I can use it.
And of course hack it, and hook my modules to it!
I'm dealing with content-transformation software, for automated markup
repairs and transformation (c.f. the Site Valet for my work, and the
WAI/ER for context). The software will need to work both in a server
and a proxy supporting HTTP/1.1 cache control. The path of least effort
for this would probably be something like IBM Websphere, but I'd much
rather base it on opensource, so I'm envisaging a new Apache module.
I know I'll need Apache 2.0, because prefork would mean jumping through
hoops to pool database connections. I know I need an HTTP/1.1 mod_proxy,
which doesn't exist (last time I looked). So yes, I'll hack mod_proxy as
necessary, and feed my work back to apache.org provided there's a suitable
mechanism for doing so. Regardless of whether it's part of the core
distribution or semi-detached.
</soapbox>
--
Nick Kew
Re: Punt the Proxy
Posted by Ben Hyde <bh...@pobox.com>.
I found Greg's note most amusing.
Greg Stein <gs...@lyra.org> writes:
> The number of end users is totally irrelevant. It is all about the people
> *working* on the code. We are not here to satisfy end users' entire scope of
> needs. We are here to satisfy them where their needs intersect our desires.
A slight tunning. We are also here to enable users, or thier agents, work
on what they desire.
I'm certain there are people out there that really really desire proxy.
They want it bundled in the server - so the trick is to structure things
so those people get traction on the problem. i.e. so they can hack, not whinge,
at it.
- ben
Re: Punt the Proxy
Posted by Eli Marmor <ma...@elmar.co.il>.
Greg Stein wrote:
> > new-httpd. In fact, I have copied this message to new-httpd as well, in
> > the hopes that people will keep the conversation on both lists.
>
> [ this is actually quite hard to do because of the Reply-To munging ]
Heh!
I'm lucky to have the honor to be subscribed to both lists ;-)
> Exactly. And we bring up tossing the proxy, and people jump back in for a
> week, then go away and proxy languishes again. Chuck has now said "I've got
> some patches!". They'll get applied, and we won't see him again for several
> months (just like last time).
I want to be practical. Let me, please, analyze the status and suggest
a compromise that will be good for all sides ("win-win"). Let's inspect
the problems one by one:
1. mod_proxy is not compiled under 2.0:
It was already OK with 2.0a8 (IIRC), and with Chuck patches it is
being fixed again (maybe it's already OK; I must admit that I didn't
have chance to check his patches).
2. mod_proxy isn't running correctly / functionality doesn't cover the
1.3 status:
See #1 above.
3. It is not maintained / there is no maintainer:
Well, YESTERDAY, you were right. CURRENTLY, *TODAY*, it is
maintained. You can't suggest the axing today. Tomorrow? Maybe. If
it is important for you that it will be maintained, then you must be
be glad that your "threats" caused a shock for some people, and they
are (re)starting to volunteer.
4. People appear for a week, out of the blue, after a threat from you,
but disappear for 2-3 months, and then appear back:
Oh! Finally we get to the real problem. So here is my suggestion (a
compromise that should make everybody happy):
Instead of immediately axing the proxy, just chance the frequency of
the threats from once per 6 months, to once per a month, or even bi-
weekly. Two scenarios may happen then:
a. The proxy users/lovers/maintainers/programmers are in a permanent
shock, and always maintain and support it. This is the best for
everybody, I believe. And it resolves problem #4.
b. Otherwise, if your "threats" don't wake people up, just realize
your threat and axe it (after voting, I guess). It looks as the
dream of some people here, isn't it? ;-)
5. The beta must be launched, but the proxy is not yet ready:
Well, first of all I'm not sure that it will not be ready in 2-3
days; Now, with Chuck, Philippe, and others, investing efforts in
that, there may be a surprise.
But even if the big day arrives, and the proxy is not yet ready,
then you may make it "experimental" (i.e. it already has its own
directory, and it doesn't make sense to move it, but it may become a
module which is compiled in *ONLY* if the user/configurator asks for
it explicitly, or in other words - it is excluded even from the
"most" option. In addition, you may add warning. Once Apache 2.0beta
is shipped with such a module, I'm convinced many people will make
their best to make it better; After all, this is one of the purposes
of having a beta, isn't it? This is the key point here; If we really
want to make mod_proxy working and stable, the best chance to do it
is by including it in a stuff that is going to be very widely
spread).
6. Some of mod_proxy functionality is duplicated by squid:
Here, my compromise doesn't bridge the arguing opinions. However, I
must disagree here: First, it's only "some" functionality. Second,
almost any part of Apache is duplicated by another Open-Source
Software. Third, licenses are not compatible. Fourth, there are
already many users who depend on mod_proxy functionality, and most
can't just migrate to squid with no pain. But the fifth one, is the
most important: Sometimes, the combination of mod_proxy with other
features of Apache, is required. In better cases, we can just add
another tier, by chaining squid to an existing Apache (ending up
with much more overhead), but in most cases, it's irrelevant. And
with the filtered I/O, many new possibilities and tricks become
possible by the combination of Apache and mod_proxy. I believe that
Apache 2.0 may become a platform for new applications and
transformations, based on this combination.
In any case, mod_proxy doesn't compete against squid, and is not
"better" or "worse" than it; These are just two different products,
that only their very basic functionality (yes, including the reverse
proxy) is similar.
So here is again my compromised suggestion: Wait a few days or a few
weeks, and see what's happenning with mod_proxy. Once a few days /
weeks paased with no serious maintenance, re-"threat", and if nobody
responds, vote for axing it. If, meanwhile, Apache 2.0 beta must be
announced and mod_proxy is not in a stable status, follow #5 above.
--
Eli Marmor
marmor@netmask.it
CTO, Founder
Netmask (El-Mar) Internet Technologies Ltd.
__________________________________________________________
Tel.: +972-9-766-1020 8 Yad-Harutzim St.
Fax.: +972-9-766-1314 P.O.B. 7004
Mobile: +972-50-23-7338 Kfar-Saba 44641, Israel
Re: Punt the Proxy
Posted by Eli Marmor <ma...@elmar.co.il>.
Greg Stein wrote:
> > new-httpd. In fact, I have copied this message to new-httpd as well, in
> > the hopes that people will keep the conversation on both lists.
>
> [ this is actually quite hard to do because of the Reply-To munging ]
Heh!
I'm lucky to have the honor to be subscribed to both lists ;-)
> Exactly. And we bring up tossing the proxy, and people jump back in for a
> week, then go away and proxy languishes again. Chuck has now said "I've got
> some patches!". They'll get applied, and we won't see him again for several
> months (just like last time).
I want to be practical. Let me, please, analyze the status and suggest
a compromise that will be good for all sides ("win-win"). Let's inspect
the problems one by one:
1. mod_proxy is not compiled under 2.0:
It was already OK with 2.0a8 (IIRC), and with Chuck patches it is
being fixed again (maybe it's already OK; I must admit that I didn't
have chance to check his patches).
2. mod_proxy isn't running correctly / functionality doesn't cover the
1.3 status:
See #1 above.
3. It is not maintained / there is no maintainer:
Well, YESTERDAY, you were right. CURRENTLY, *TODAY*, it is
maintained. You can't suggest the axing today. Tomorrow? Maybe. If
it is important for you that it will be maintained, then you must be
be glad that your "threats" caused a shock for some people, and they
are (re)starting to volunteer.
4. People appear for a week, out of the blue, after a threat from you,
but disappear for 2-3 months, and then appear back:
Oh! Finally we get to the real problem. So here is my suggestion (a
compromise that should make everybody happy):
Instead of immediately axing the proxy, just chance the frequency of
the threats from once per 6 months, to once per a month, or even bi-
weekly. Two scenarios may happen then:
a. The proxy users/lovers/maintainers/programmers are in a permanent
shock, and always maintain and support it. This is the best for
everybody, I believe. And it resolves problem #4.
b. Otherwise, if your "threats" don't wake people up, just realize
your threat and axe it (after voting, I guess). It looks as the
dream of some people here, isn't it? ;-)
5. The beta must be launched, but the proxy is not yet ready:
Well, first of all I'm not sure that it will not be ready in 2-3
days; Now, with Chuck, Philippe, and others, investing efforts in
that, there may be a surprise.
But even if the big day arrives, and the proxy is not yet ready,
then you may make it "experimental" (i.e. it already has its own
directory, and it doesn't make sense to move it, but it may become a
module which is compiled in *ONLY* if the user/configurator asks for
it explicitly, or in other words - it is excluded even from the
"most" option. In addition, you may add warning. Once Apache 2.0beta
is shipped with such a module, I'm convinced many people will make
their best to make it better; After all, this is one of the purposes
of having a beta, isn't it? This is the key point here; If we really
want to make mod_proxy working and stable, the best chance to do it
is by including it in a stuff that is going to be very widely
spread).
6. Some of mod_proxy functionality is duplicated by squid:
Here, my compromise doesn't bridge the arguing opinions. However, I
must disagree here: First, it's only "some" functionality. Second,
almost any part of Apache is duplicated by another Open-Source
Software. Third, licenses are not compatible. Fourth, there are
already many users who depend on mod_proxy functionality, and most
can't just migrate to squid with no pain. But the fifth one, is the
most important: Sometimes, the combination of mod_proxy with other
features of Apache, is required. In better cases, we can just add
another tier, by chaining squid to an existing Apache (ending up
with much more overhead), but in most cases, it's irrelevant. And
with the filtered I/O, many new possibilities and tricks become
possible by the combination of Apache and mod_proxy. I believe that
Apache 2.0 may become a platform for new applications and
transformations, based on this combination.
In any case, mod_proxy doesn't compete against squid, and is not
"better" or "worse" than it; These are just two different products,
that only their very basic functionality (yes, including the reverse
proxy) is similar.
So here is again my compromised suggestion: Wait a few days or a few
weeks, and see what's happenning with mod_proxy. Once a few days /
weeks paased with no serious maintenance, re-"threat", and if nobody
responds, vote for axing it. If, meanwhile, Apache 2.0 beta must be
announced and mod_proxy is not in a stable status, follow #5 above.
--
Eli Marmor
marmor@netmask.it
CTO, Founder
Netmask (El-Mar) Internet Technologies Ltd.
__________________________________________________________
Tel.: +972-9-766-1020 8 Yad-Harutzim St.
Fax.: +972-9-766-1314 P.O.B. 7004
Mobile: +972-50-23-7338 Kfar-Saba 44641, Israel
Re: Punt the Proxy
Posted by Greg Stein <gs...@lyra.org>.
On Tue, Feb 06, 2001 at 11:36:59AM -0800, rbb@covalent.net wrote:
> Eli Marmor wrote:
> >
> > I hope I didn't sound too much flaming...
>
> I didn't take the message as a flame at all. I just wanted to clarify the
> position. You are correct, that this needs to be posted here as well as
> new-httpd. In fact, I have copied this message to new-httpd as well, in
> the hopes that people will keep the conversation on both lists.
[ this is actually quite hard to do because of the Reply-To munging ]
> > I remember at least 3 different people (don't remember their names; I
> > think that one of them was from IBM), volunteered to maintain it. I
> > guess that all of them are subscribed to this list.
>
> This is the problem. People have volunteered to be the maintainer of the
> proxy for years, but it never really happens.
Exactly. And we bring up tossing the proxy, and people jump back in for a
week, then go away and proxy languishes again. Chuck has now said "I've got
some patches!". They'll get applied, and we won't see him again for several
months (just like last time).
This is bogus, and we should stop deluding ourselves that proxy is being
maintained. Hell, that it is even getting finished.
>...
> By removing the proxy from the main tree, we remove the second problem,
> allowing more work to happen on the proxy. If we continue to distribute
> the proxy from the web site, I do not believe that we will decrease the
> number of people using it.
Right. If people need the proxy functionality, then they know it is on
apache.org; they grab it and install it. No big deal.
> > I believe that proxy must stay in the standard tree. If the problem is
> > that it is not up-to-date, then update it. If the problem is that it's
> > broken, then fix it. If the problem is that there is no maintainer,
> > then find one. But axing it, is like axing other integral parts of
Oh, shut up. We aren't responsible for the proxy code. Every works on what
they want to work on. We can't point at somebody and say "you are now the
maintainer." That just doesn't work in an Open Source project.
We'll move the code into a new CVS tree, and then YOU work on it. Don't come
crying to us because your favorite feature is being moved to a different
tree. We aren't here to please you, we're here to work on things that are
interesting to us.
If you want it, then step up to the plate and contribute your time. If you
don't want to spend your own time on it, then just go away.
>...
> I believe many of those "fixes" are actually incorrect in Apache 2.0. I
> have been investigating them slowly, but it should be 100% possible to
> create a fully independant proxy. If it isn't, then I would like to know
> why and fix those problems.
I'm with Ryan. If the proxy cannot be built as a separate entity, then we
need to refactor the core to make it possible. But the core does not
necessarily need to explicitly deal with proxy.
> > I just think about the MANY existing users of it. I'm sure there are
> > more users of Apache's proxy, than users of BeOS or NetWare, that you
> > are working so hard to let them use Apache.
This has absolutely no relevance whatsoever. BeOS and Netware users are
taking the time to make Apache work on their platform. Nobody is doing that
for proxy.
The number of end users is totally irrelevant. It is all about the people
*working* on the code. We are not here to satisfy end users' entire scope of
needs. We are here to satisfy them where their needs intersect our desires.
>...
> Eli, I truly think you and I are working for the same thing. I want to
> see the proxy work continue. I believe by removing the proxy from the
> tree, we can make it much easier for more people to contribute to the
> proxy. I also honestly believe that if we can prove the proxy has a
> healthy development team behind it, then in the future it should be very
> possible to fold the new proxy back into the main tree.
I totally agree. Move the proxy out. Let people work on it and get it
functional. Then make a decision at that point to include it into the tree
or to keep the development separate.
But keeping it in the tree hasn't work for at least a year, probably two.
Enough is enough. Move it out.
Cheers,
-g
--
Greg Stein, http://www.lyra.org/
Re: Punt the Proxy
Posted by Greg Stein <gs...@lyra.org>.
On Tue, Feb 06, 2001 at 11:36:59AM -0800, rbb@covalent.net wrote:
> Eli Marmor wrote:
> >
> > I hope I didn't sound too much flaming...
>
> I didn't take the message as a flame at all. I just wanted to clarify the
> position. You are correct, that this needs to be posted here as well as
> new-httpd. In fact, I have copied this message to new-httpd as well, in
> the hopes that people will keep the conversation on both lists.
[ this is actually quite hard to do because of the Reply-To munging ]
> > I remember at least 3 different people (don't remember their names; I
> > think that one of them was from IBM), volunteered to maintain it. I
> > guess that all of them are subscribed to this list.
>
> This is the problem. People have volunteered to be the maintainer of the
> proxy for years, but it never really happens.
Exactly. And we bring up tossing the proxy, and people jump back in for a
week, then go away and proxy languishes again. Chuck has now said "I've got
some patches!". They'll get applied, and we won't see him again for several
months (just like last time).
This is bogus, and we should stop deluding ourselves that proxy is being
maintained. Hell, that it is even getting finished.
>...
> By removing the proxy from the main tree, we remove the second problem,
> allowing more work to happen on the proxy. If we continue to distribute
> the proxy from the web site, I do not believe that we will decrease the
> number of people using it.
Right. If people need the proxy functionality, then they know it is on
apache.org; they grab it and install it. No big deal.
> > I believe that proxy must stay in the standard tree. If the problem is
> > that it is not up-to-date, then update it. If the problem is that it's
> > broken, then fix it. If the problem is that there is no maintainer,
> > then find one. But axing it, is like axing other integral parts of
Oh, shut up. We aren't responsible for the proxy code. Every works on what
they want to work on. We can't point at somebody and say "you are now the
maintainer." That just doesn't work in an Open Source project.
We'll move the code into a new CVS tree, and then YOU work on it. Don't come
crying to us because your favorite feature is being moved to a different
tree. We aren't here to please you, we're here to work on things that are
interesting to us.
If you want it, then step up to the plate and contribute your time. If you
don't want to spend your own time on it, then just go away.
>...
> I believe many of those "fixes" are actually incorrect in Apache 2.0. I
> have been investigating them slowly, but it should be 100% possible to
> create a fully independant proxy. If it isn't, then I would like to know
> why and fix those problems.
I'm with Ryan. If the proxy cannot be built as a separate entity, then we
need to refactor the core to make it possible. But the core does not
necessarily need to explicitly deal with proxy.
> > I just think about the MANY existing users of it. I'm sure there are
> > more users of Apache's proxy, than users of BeOS or NetWare, that you
> > are working so hard to let them use Apache.
This has absolutely no relevance whatsoever. BeOS and Netware users are
taking the time to make Apache work on their platform. Nobody is doing that
for proxy.
The number of end users is totally irrelevant. It is all about the people
*working* on the code. We are not here to satisfy end users' entire scope of
needs. We are here to satisfy them where their needs intersect our desires.
>...
> Eli, I truly think you and I are working for the same thing. I want to
> see the proxy work continue. I believe by removing the proxy from the
> tree, we can make it much easier for more people to contribute to the
> proxy. I also honestly believe that if we can prove the proxy has a
> healthy development team behind it, then in the future it should be very
> possible to fold the new proxy back into the main tree.
I totally agree. Move the proxy out. Let people work on it and get it
functional. Then make a decision at that point to include it into the tree
or to keep the development separate.
But keeping it in the tree hasn't work for at least a year, probably two.
Enough is enough. Move it out.
Cheers,
-g
--
Greg Stein, http://www.lyra.org/
Re: Punt the Proxy
Posted by rb...@covalent.net.
> I hope I didn't sound too much flaming...
I didn't take the message as a flame at all. I just wanted to clarify the
position. You are correct, that this needs to be posted here as well as
new-httpd. In fact, I have copied this message to new-httpd as well, in
the hopes that people will keep the conversation on both lists.
> I remember at least 3 different people (don't remember their names; I
> think that one of them was from IBM), volunteered to maintain it. I
> guess that all of them are subscribed to this list.
This is the problem. People have volunteered to be the maintainer of the
proxy for years, but it never really happens. I honestly believe that a
part of the reason for that is because that person must not only deal with
technical arguments on this list (those are 100% valid and they help the
code), but they must also deal with the people on new-httpd who don't see
the need for the proxy in the main tree.
By removing the proxy from the main tree, we remove the second problem,
allowing more work to happen on the proxy. If we continue to distribute
the proxy from the web site, I do not believe that we will decrease the
number of people using it.
> I believe that proxy must stay in the standard tree. If the problem is
> that it is not up-to-date, then update it. If the problem is that it's
> broken, then fix it. If the problem is that there is no maintainer,
> then find one. But axing it, is like axing other integral parts of
There is the rub. Fixing it is up to the maintainer. The problem is that
currently the maintainer must be an httpd committer. By removing it from
the httpd tree, we can lighten that load, so that more people can easily
get commit access to the proxy and it won't fall on one person's shoulders
to maintain the proxy.
> Apache. I don't want to list the reasons behind this claim, because
> there are many in this list that will do it better than me. But just
> think about one thing: There are many fixes to the core tree of Apache
> to allow proxy to run. Like proxyreq. And zillion of other things. And
> if you are going to include the reverse proxy in any case, the
> difference is not too high. After all, one may compile Apache without
> the proxy stuff, so the weight of it is not relevant.
I believe many of those "fixes" are actually incorrect in Apache 2.0. I
have been investigating them slowly, but it should be 100% possible to
create a fully independant proxy. If it isn't, then I would like to know
why and fix those problems.
> I just think about the MANY existing users of it. I'm sure there are
> more users of Apache's proxy, than users of BeOS or NetWare, that you
> are working so hard to let them use Apache. This is also why I started
> the "war" between Ben and Ralf regarding SSL (well, now they are
> fighting each other, but it's much better than the status before, when
> nothing was done...). It's pity that Apache 2.0 is so close, and the
> work on SSL even didn't started. There are so many users of SSL. If the
> users of BeOS and NetWare and EBCDIC are so important for us, the 50,000%
> more users of the proxy and 100,000% more of SSL (I threw numbers without
> checking ;-), are VERY important. I don't believe that everything should
> be included in the tree (somebody suggested to include mod_macro or
> mod_layout, and I must admit that it sounded bad for me), but axing proxy
> or mod_rewrite is not looking good for me.
Eli, I truly think you and I are working for the same thing. I want to
see the proxy work continue. I believe by removing the proxy from the
tree, we can make it much easier for more people to contribute to the
proxy. I also honestly believe that if we can prove the proxy has a
healthy development team behind it, then in the future it should be very
possible to fold the new proxy back into the main tree.
I hope I have expressed my hopes for this well. I am not at all
interested in seeing the proxy dropped from Apache. I am 100% behind
making it much easier for people to help with the proxy.
Ryan
_______________________________________________________________________________
Ryan Bloom rbb@apache.org
406 29th St.
San Francisco, CA 94131
-------------------------------------------------------------------------------
Re: Punt the Proxy
Posted by rb...@covalent.net.
> I hope I didn't sound too much flaming...
I didn't take the message as a flame at all. I just wanted to clarify the
position. You are correct, that this needs to be posted here as well as
new-httpd. In fact, I have copied this message to new-httpd as well, in
the hopes that people will keep the conversation on both lists.
> I remember at least 3 different people (don't remember their names; I
> think that one of them was from IBM), volunteered to maintain it. I
> guess that all of them are subscribed to this list.
This is the problem. People have volunteered to be the maintainer of the
proxy for years, but it never really happens. I honestly believe that a
part of the reason for that is because that person must not only deal with
technical arguments on this list (those are 100% valid and they help the
code), but they must also deal with the people on new-httpd who don't see
the need for the proxy in the main tree.
By removing the proxy from the main tree, we remove the second problem,
allowing more work to happen on the proxy. If we continue to distribute
the proxy from the web site, I do not believe that we will decrease the
number of people using it.
> I believe that proxy must stay in the standard tree. If the problem is
> that it is not up-to-date, then update it. If the problem is that it's
> broken, then fix it. If the problem is that there is no maintainer,
> then find one. But axing it, is like axing other integral parts of
There is the rub. Fixing it is up to the maintainer. The problem is that
currently the maintainer must be an httpd committer. By removing it from
the httpd tree, we can lighten that load, so that more people can easily
get commit access to the proxy and it won't fall on one person's shoulders
to maintain the proxy.
> Apache. I don't want to list the reasons behind this claim, because
> there are many in this list that will do it better than me. But just
> think about one thing: There are many fixes to the core tree of Apache
> to allow proxy to run. Like proxyreq. And zillion of other things. And
> if you are going to include the reverse proxy in any case, the
> difference is not too high. After all, one may compile Apache without
> the proxy stuff, so the weight of it is not relevant.
I believe many of those "fixes" are actually incorrect in Apache 2.0. I
have been investigating them slowly, but it should be 100% possible to
create a fully independant proxy. If it isn't, then I would like to know
why and fix those problems.
> I just think about the MANY existing users of it. I'm sure there are
> more users of Apache's proxy, than users of BeOS or NetWare, that you
> are working so hard to let them use Apache. This is also why I started
> the "war" between Ben and Ralf regarding SSL (well, now they are
> fighting each other, but it's much better than the status before, when
> nothing was done...). It's pity that Apache 2.0 is so close, and the
> work on SSL even didn't started. There are so many users of SSL. If the
> users of BeOS and NetWare and EBCDIC are so important for us, the 50,000%
> more users of the proxy and 100,000% more of SSL (I threw numbers without
> checking ;-), are VERY important. I don't believe that everything should
> be included in the tree (somebody suggested to include mod_macro or
> mod_layout, and I must admit that it sounded bad for me), but axing proxy
> or mod_rewrite is not looking good for me.
Eli, I truly think you and I are working for the same thing. I want to
see the proxy work continue. I believe by removing the proxy from the
tree, we can make it much easier for more people to contribute to the
proxy. I also honestly believe that if we can prove the proxy has a
healthy development team behind it, then in the future it should be very
possible to fold the new proxy back into the main tree.
I hope I have expressed my hopes for this well. I am not at all
interested in seeing the proxy dropped from Apache. I am 100% behind
making it much easier for people to help with the proxy.
Ryan
_______________________________________________________________________________
Ryan Bloom rbb@apache.org
406 29th St.
San Francisco, CA 94131
-------------------------------------------------------------------------------
Re: Punt the Proxy
Posted by Eli Marmor <ma...@elmar.co.il>.
I hope I didn't sound too much flaming...
Anyway, I must admit, Ryan, that the effort invested in Apache 2.0, is
really impressive and great, and your part in particular. You are the
real hero of Apache 2.0.
I also wrote in my original message that Apache 2.0 is going to be a
great thing.
And I also wrote that the most important thing is that somebody will
become the official maintainer of the proxy.
I remember at least 3 different people (don't remember their names; I
think that one of them was from IBM), volunteered to maintain it. I
guess that all of them are subscribed to this list.
Unfortunately, I don't think they are subscribed to new-httpd too. So
raising the issue in new-httpd, as you and others did, was not enough.
It was important to do it here too, and in a provocative way as I used
(lakonic message would not be enough to wake the people up).
I believe that proxy must stay in the standard tree. If the problem is
that it is not up-to-date, then update it. If the problem is that it's
broken, then fix it. If the problem is that there is no maintainer,
then find one. But axing it, is like axing other integral parts of
Apache. I don't want to list the reasons behind this claim, because
there are many in this list that will do it better than me. But just
think about one thing: There are many fixes to the core tree of Apache
to allow proxy to run. Like proxyreq. And zillion of other things. And
if you are going to include the reverse proxy in any case, the
difference is not too high. After all, one may compile Apache without
the proxy stuff, so the weight of it is not relevant.
I just think about the MANY existing users of it. I'm sure there are
more users of Apache's proxy, than users of BeOS or NetWare, that you
are working so hard to let them use Apache. This is also why I started
the "war" between Ben and Ralf regarding SSL (well, now they are
fighting each other, but it's much better than the status before, when
nothing was done...). It's pity that Apache 2.0 is so close, and the
work on SSL even didn't started. There are so many users of SSL. If the
users of BeOS and NetWare and EBCDIC are so important for us, the 50,000%
more users of the proxy and 100,000% more of SSL (I threw numbers without
checking ;-), are VERY important. I don't believe that everything should
be included in the tree (somebody suggested to include mod_macro or
mod_layout, and I must admit that it sounded bad for me), but axing proxy
or mod_rewrite is not looking good for me.
--
Eli Marmor
marmor@netmask.it
CTO, Founder
Netmask (El-Mar) Internet Technologies Ltd.
__________________________________________________________
Tel.: +972-9-766-1020 8 Yad-Harutzim St.
Fax.: +972-9-766-1314 P.O.B. 7004
Mobile: +972-50-23-7338 Kfar-Saba 44641, Israel
Re: Punt the Proxy
Posted by rb...@covalent.net.
I should probably comment on this. I am the person who suggesting
removing the proxy from the tree. I am also one of the people who did the
port to 2.0, just to prove that I am not against the proxy.
What I am against, is having a bunch of code in the tree that doesn't
compile when it is downloaded, and hasn't been worked on in ages.
The current proposal doesn't kill the proxy, it removes it with an option
to add it back. The idea is that the proxy will be moved off to another
repository, where people who want to work on it can easily. All comments
can be sent here, and the CVS tree would have it's own mailing list as
well. When the code becomes stable and there is an active maintainer,
somebody who makes sure the code always compiles, the proxy may get back
into the httpd-2.0 tree.
If it doesn't, the proxy would be a sub-project of httpd, and it would
have a prominant spot on the httpd.apache.org web page.
This is just a way to shink the size of the httpd tarball, by removing
some code that isn't currently being maintained.
Ryan
On Tue, 6 Feb 2001, Eli Marmor wrote:
> Hi list!
>
> Are you aware of the discussion at new-httpd@apache.org to punt the
> Proxy from the source tree?
>
> Very strong opinions are expressed there, against mod_proxy as a part
> of an HTTP server, against mod_proxy as a module without a maintainer,
> etc., etc.
>
> I believe that most of us opposed these opinions. On the other hand,
> most of us don't have time to join the main list and receive the
> zillion messages, so are not aware of that discussion, and almost
> nobody tries to argue with the opinions there.
>
> I welcome subscribers of modproxy-dev to join that list (instruction
> at http://www.apache.org/foundation/mailinglists.html#http-dev) and
> express our side. To see the opinions I mentioned, please look at the
> following threads:
>
> http://marc.theaimsgroup.com/?t=97595977400003&w=2&r=1
> http://marc.theaimsgroup.com/?l=apache-new-httpd&m=98140196311131&w=2
>
> The best thing we can contribute to the discussion, is somebody who
> will take the responsibility to be the official maintainer of this
> module.
>
> But even if you can't do it, you can still:
>
> 1. Write about the importance of the module.
> 2. Emphasize the extra importance of the reverse proxy, and that if it
> is included, it will make sense to include the whole proxy.
> 3. Features (including ease, weight, rewrite, etc.) which are not
> supported by squid but only by Apache.
> 4. your experience with it (e.g. "I'm the webmaster of ...... and we
> use mod_proxy under 200,000 machines to cache the traffic of our
> 120 million users. Axing mod_proxy from the source tree, will not
> only cause us a problem, but also lower its netcraft statistics
> from 60% to 2%").
> 5. Try to convince that today, with the new mechanism of filtered I/O
> in Apache 2.0, mod_proxy is an integral part of Apache more than
> ever. IIRC, one of the subscribers here, told about his experience
> with porting the proxy to use the filters API, and it sounded
> great.
> 6. Send patches.
> 7. Tell them about the great success that Apache 2 is going to have:
> After 4 years of development, and without key parts such as proxy
> and SSL (now even mod_rewrite is under danger!), I'm convinced that
> Apache 2 will double netcraft statistics (from 60% to 120%! ;-)
>
> Now seriously (contrary to 7): Apache 2.0 is really going to be a
> great thing; It will be pity if the key parts of it will be missing.
> So open your mouth and speak.
> --
> Eli Marmor
> marmor@netmask.it
> CTO, Founder
> Netmask (El-Mar) Internet Technologies Ltd.
> __________________________________________________________
> Tel.: +972-9-766-1020 8 Yad-Harutzim St.
> Fax.: +972-9-766-1314 P.O.B. 7004
> Mobile: +972-50-23-7338 Kfar-Saba 44641, Israel
>
>
_______________________________________________________________________________
Ryan Bloom rbb@apache.org
406 29th St.
San Francisco, CA 94131
-------------------------------------------------------------------------------