httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From TOKI...@aol.com
Subject Re: zlib inclusion and mod_gz(ip) recap
Date Sat, 08 Sep 2001 00:59:24 GMT

> > Ken Coar wrote...
> > Current consensus appears to be to add it to modules/experimental.
>  
>  Ryan Bloom responded...
>
>  I don't see how that could possibly be the consensus, since I have -1 in
>  the STATUS file.

The 'consensus' that Ken thinks he sees just diminished.

I am (obviously) not a committer but on the hopes that my opinion
might actually mean something... I am changing my 'vote' on +1
to adding mod_gz following the broil fest to a -1 so now there are
at least 2 definite 'Not at this time' votes.

Here is why...

Regardless of the fact that I might be the author of any publicly
available software that provides some similar functionality something 
is just plain screwy about the way this mod_gz thing is being (seemingly)
bombed into the core.

If I knew nothing about compression and/or Apache modules and/or
the current state of 2.0 I would still have to say something is weird
here and doesn't seem to following the usual *careful* path for
putting anything into an Apache core distribution.

Here are just a few of the things that ( IMHO ) are a little 'weird'
with regards to this all suddenly being a house on fire...

1. The very self-described 'newbie' (Justin) who came out of nowhere 
and basically said 'Hey fellas... I haven't load tested this mod_gz thing 
but it worked with my Netscape browser ( only one tested? ) once or
twice so let's bomb Ian's mod_gz into the core right now' now says 
he is going AWAY for 2 weeks.

The very guy who wanted it there IMMEDIATELY is now disappearing
for 2 weeks? That's odd.

Granted... Ian has done some more testing since Justin's request 
( I think? ) and posted 1 page of results ( Nothing following that ) but 
at the time the request for core insertion was made there was no evidence 
of any testing at all except the most basic of scenarios ( Ask for an
HTML page and watch it appear in a browser ).

2. I asked Justin myself where the 'fire' was and what prompted
him to request the core-bomb-in and although he gave a fair 
answer as to 'it ought to be there' ( which I cannot disagree with )
he didn't really establish much of a basis for actual Apache CORE 
insertion AT THIS MOMENT IN TIME.

If ( as he said ) he has seen 'weird things' while testing mod_include
and mod_gz and ( as he said ) feels there needs to be 'another filter
involved' for good testing... then the insertion of Ian's demo into the core
AT THIS MOMENT is not a pre-requisite to debugging something. 
Ian's code has been posted and it's there. If anyone needs to use it in 
their 2.0 testing well... then just go ahead and DO it. 

That process ( just using it right away as-is ) would help find these
filtering problems Justin seemed to be talking about AND I suppose
it would help debug mod_gz itself BEFORE it becomes permanently
part of anything.

I still don't see why the actual insertion of a not-very-well-tested 
filtering demo into the core and the legal/official process that generates 
is somehow a pre-requisite to finding filtering 'problems' in pre-BETA 2.0.

3. Ian doesn't even understand why Jim J. would ask for a signed
module submission form. No offense to Ian but this doesn't indicate
that he himself understands the official module submission process
and the legalities that surround it. Certainly not a showstopper but
just part of the 'weirdness'. We got blasted all weekend for 'not 
knowing how Apache works'... No one needs to blast Ian but he
certainly needs to understand better why he must give signed 
consent for his work.

4. It seems no one has even proof-read the actual mod_gz code much.
For starters... there are places where it's going to bomb simply because 
it's using integers instead of longs. It won't show up until something
is over 65k. It also doesn't really have the right ZLIB copyright information
in it which would normally be the 'correct' thing for Apache to do with
any public-domain dependent code dropped into the core.

5. Dr Mark Adler ( co-author of ZLIB itself with Jean ) is currently on
our payroll. He and I are currently discussing a few of the issues that
were raised during the broil-fest about the actual suitability of ZLIB
for your purposes. Somewhere in the discussions we tried to calmly
point out what the concerns can/should be but all that got lost in
the noise, methinks. It still should concern any Apache what the
actual implications are of including a module in the core distribution
that tries to use ZLIB as a real-time data compression engine.

The more I think about it... ZLIB is probably the way for you
guys to go even though mod_gzip doesn't use it because using
ZLIB would simply be easier for you guys in the long run. 
Someone raised the 'trust' issue. Well... if you trust ZLIB
then you should probably use it in lieu of anything else.
You guys are not in the compression business.

Ryan Bloom himself said in about the third message following
Justin's '"Let's drop this mod_gz puppy in the core" message that
perhaps waiting until we hear from some people who are 'not on
the list' was a good idea. Neither myself nor Peter had even joined
the discussion at that point and Ryan's instincts were correct.

Maybe you should wait just a little longer to 'hear' from Mark ( Adler )
himself at this point. If you don't trust what I know about ZLIB then
I think you should trust what HE (Mark) has to say.

Ok... enuff said...

I am sure someone is going to come back with 'Methinks he doth
protest too much' and/or some are going to do the usual 'iterate the 
points' and comment back but it's really not necessary. All I was doing 
is iterating MY reasons why I have changed MY earlier +1 to a -1.

People are free to stick with whatever negative or positive integer
they want... mine has gone to -1 on this, at least for the moment.

Hard as it might seem for some of you who have a fixed-in-stone
opinion about RC and/or me personally to believe... I still
think this is a pretty good HTTP Server and my instincts still
require me to speak up when I believe I see a mistake being made.

It isn't going to hurt anything to wait a little while ( perhaps even
simply until the original advocate of the insertion gets back from the
2 week trip? ) and then just pick it up again. Maybe mod_gz will
be even better ( and more well tested ) by then.

That's all.
Take it FWIW.
Thanks for your time.

Kevin Kiley





Mime
View raw message