Re: mod_rewrite

[Marc Slemko <ma...@znep.com>]

On Thu, 30 Jan 1997, Ralf S. Engelschall wrote:

> > Having said all that, I don't think that the Apache group can be expected
> > to track down every contributed module author to sync changes with
> > their own release cycle. This whole release thing is getting tough enough
> > to pull off.
> 
> > I hope that Ralf can help us get things in sync. Do we have concerns
> > about some of the new features of his latest version?
> 
> I always try to keep mod_rewrite in sync, while actually this should never be
> needed. Because mod_rewrite is for Apache and for the Apache Group. I have no
> plans to do an own thing. The only difference is that I release newer version
> before they are intergrated into the Apache distribution. 

No.  The difference is also that since they are two seperate development
efforts, no matter how they may be integrated at the end, they have two
different development schedules.

> 
> The problem now is that the Apache Group commited the patches directly to
> their CVS without any notice by me. I never even noticed that mod_rewrite was
> also patched for snprintf. So 1.2b6 went out with a broken mod_rewrite! It is
> ok and really fine when there are any ideas or notes or patches for
> mod_rewrite from the Apache Group.  But because I think I'm the maintainer of
> mod_rewrite it would be very nice if the Apache Group would wait for a +1 of
> me for any patch to mod_rewrite. Or even better: Let me include the change
> directly to the official mod_rewrite release and then commit this version
> again into the Apache CVS.

You think it is reasonable for someone to jump upon you mentioning
a new version once, yet you think it unreasonable for you to notice 
a patch that was posted at least three times in addition to the
commit message, was discussed at length as changing a large percent of
the files and which I explicitly asked anyone with any particular
interest in any code to review that bit?

There was a good month for you to comment on them.  You didn't.

> I cannot trace every day the CVS commit messages and re-integrate the changes
> into the official release. Because often (as with ap_snprintf) the patched

If you want "The Apache Group" to trace your versions and reintegrate
the changes into Apache (which, in this case, mainly means reviewing
the differences and testing) you have to make some effort.  It isn't
that hard to setup a mail filter that will alert you to any patch or
commit message about mod_rewrite.

> version is no longer with the spirit of the original, i.e.  compilable under
> _BOTH_ under Apache 1.1.1 _AND_  Apache 1.2bX. So, I got the trouble and
> bugreports and the Apache Group had a broken 1.2b6 :-( Hmmmm...

Apache is distributed with a version that works with the current
source tree.  If you want the code to work with any particular other
version, that is really your responsibility.

> I would really prefer the Apache Group would privide me with core membership
> and CVS access.  Then I can vote for the mod_rewrite changes _AND_ do the
> commitment myself. When it is then broken, I can fix it immediately. But this
> time we have to wait for Marc S. because he has introduced the bug and I
> remembered it too late. So, my question is: Are there any problems to add me
> to the core members? If not, please do it. Then things can go more smoothly
> and without such trouble. 

Just because you wrote the code doesn't mean you can dictate what
changes should be made when.  Apache can not be developed if for every
second change people have to specially contact this person or that
person to get their blessing.  Trying to say that this bug would have
been prevented if you had commit access does not make sense because you
have the same ability now to comment on patches as you would if you
had commit access.  When a patch like that changes several modules it
is not practical or desirable to break it down into numerous commits
simply because someone has responsibility for some particular bit of
code.

> 
> I see the days of Apache 2.0 and mod_rewrite 2.5 where we run into the same
> problem with different versions if we don't care or find a better way to sync.
> Hmmmm...
> 
> PLEASE THINK ABOUT ADDING ME TO THE CORE TEAM TO MAINTAIN MOD_REWRITE!

I won't comment on giving you commit access, since I'm not really
qualified to (other than to say I think any increase in the number of
Apache developers is good), but I will say that it is possible to
do things better without needing to have you commit all the changes
yourself.

Some of the things:
	- try to _synchronize_ your release schedule with that of Apache.
	- send _patches_.  patches can be reviewed, to review what is
	  essentially the re-import of an external bit of software
	  takes a _lot_ more effort because people have to do the
	  patches themself.
	- send updates more _often_.  Don't necessarily just integrate
	  the changes when Apache is getting ready for release.  No
	  matter how good your software engineering and release
	  testing may be, I don't think it is reasonable for them to
	  be blindly trusted; nothing to do with _you_, simply with
	  all externally developed software.  Currently mod_rewrite
	  isn't being developed with Apache, it is simply an external
	  bit of software that is being reimported on occasion.  
	- _pester_ people.  If you get no response, follow up on it.  

Regardless of if you have access to commit the patches yourself,
things remain roughly the same.  They still need to be reviewed
before they are committed.

On Thu, 30 Jan 1997, Ralf S. Engelschall wrote:

> Ok, but I HAVE posted my 2.4.0 announce to the list and already said that we
> should commit it for 1.2b5! But no response. Instead the group committed the
> snprintf patches and 1.2b6 was broken. Hmmm... what should I do?

Whine at the person who broke it after they broke it (which you did),
but also whine earlier when no one responds.  As pointed out by
someone, the "Apache Team" is really just the "Apache bunch of
_people_".  

> 
> >   With respect to snprintf, Ralf wants to be able to #ifdef
> > around calls to snprintf for apache 1.1.1 compatibility? I don't know
> > if that's such a good idea..
> 
> Yes, it does not look nice, so I have not done it yet. It would be better if I
> would say "2.4.1 is the latest for Apache 1.1.1" and then strip out all
> 1.1.1-#ifdefs and put in the snprintf patch and then commit it for 1.2b7.
> This would be the best this time! 
> 
> BUT THIS DOESN'T SOLVE THE PROBLEM of patches to mod_rewrite from the Apache
> Group without any +1 from me...

You have to offer your vote.  If you don't offer it, I don't consider
it reasonable to go hunting for it.