httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Bill Stoddard <stodd...@raleigh.ibm.com>
Subject Re: The way forward: 1.3.7/1.4.0/2.0/2.1/hybrid/mpm/...
Date Thu, 24 Jun 1999 22:55:56 GMT
Dean Gaudet wrote:
> 
> I'd say the only "plan" we can have for 2.0 is to see what developers feel
> like working on.
> If you want, move it to another repository.  It makes little difference to
> me -- my efforts are going into mpm... because I'm tired of people using
> apache in benchmarks and then moaning when it doesn't kick ass.  I don't
> think manoj/ryan's code is going to catch up with the leaders; nor do I
> think it makes any huge improvements over apache-1.3 as it stands. 
I mostly agree. It doesn't make huge improvements and it will not catch
up with the leaders. However, the hybrid code will support many more
concurrent clients (at least on AIX. Linux I'm not so sure since threads
and processes are basically the same there). It is a significant
improvement. Yes I know, the async mpm will support orders of magnitude
more concurrent clients.
 
> mpm's
> architecture is such that the old prefork model, plus manoj/ryan's code,
> plus my new code, plus zach brown's sigwait stuff, plus a winnt specific
> mpm, plus a win95 mpm, plus ... can all live together peacefully.
IMHO, this is a better than relying on APR to make all the core
process/thread management code common across all platforms. 
> 
> So I dunno, all I know is that I'm not sitting around waiting for a group
> consensus on anything.  I'm writing code which I want, and which I'm
> having fun writing.  I wasn't having fun waiting.  I'm also attempting to
> please as many people as possible within the constraints that I don't get
> paid for this, so if I'm not having fun then why should I do any of it?
Have fun! If we can't keep up, that's our problem.

> 
> If all we wanted to release as the next version was a threaded apache, why
> didn't we release my apache-nspr stuff 15 months ago?  It was as far from
> being stable then as apache-apr/pthread is from being stable now.
> apache-2.0/mpm is a little further from stability... but if you wanted to
> force a quick release of it all that really differs from
> apache-apr/pthread is the BUFF stuff.  And that can be made stable by just
> reverting it to 1.3 code and I'll go off and do my async stuff for apache
> 2.1.

This sounds like a plan:
1. Release 1.3.7
2. Build Apache 2.0 tree from 1.3.7, maintaining cvs history.
3. Apache 2.0 icludes:
   - mpm code restructure
   - mpms: pre-fork, win32, hybrid mpm
   - ?
4. Apache 2.1 includes:
   - mpms: async I/O
   - I/O layering
   - ?

This will let us get 2.0 out and give us more time to experiment with
I/O layering, async I/O and the fun stuff before we inflict it on our
user base :-). Sound reasonable?


-- 
Bill Stoddard
stoddard@raleigh.ibm.com

Mime
View raw message