[PATCH] 10x patches for 1.3.9

[Mike Abbott <mj...@trudge.engr.sgi.com>]

Apache Developers,

I have ported my patches (that increase Apache's performance up to
ten-fold) to Apache/1.3.9.  Only one large patch is available since I
don't have time to separate the changes into ten patches as before.
This patch also includes a bugfix in my quick CLF code (horrors! :)) and
a new feature to use direct I/O.

Rather than dump the huge patch to the mailing list I'll just point you
to it so you can download it if you're interested.  This patch and all
my old ones are available at:
	http://reality.sgi.com/mja/apache

The base for this patch is Apache/1.3.9.

As always, I welcome your review, questions, and feedback.
--
Michael J. Abbott        mja@sgi.com        http://reality.sgi.com/mja

Re: [PATCH] 10x patches for 1.3.9

[Ryan Bloom <rb...@raleigh.ibm.com>]

Just as a quick note about when 2.0 will be ready for alpha/beta.  For the
most part I agree with Manoj, we are very close.  I still have one project
that I need to finish before we announce any kind of release.  I need to
port time variables to use APR.  I want to do this before we announce an
alpha or beta, because it is a big change.  I will be working on this next
week, so it shouldn't hold us up too long now.  I also think there are
some basic features for Windows that need to be put back in.  The most
basic of these is error logs, which currently aren't working well.  I see
an alpha in early January with a beta following soon after, and an actual
release sometime in March or April at the earliest.

About the performance patches, I am -1 on putting them in for 1.3.10, -0
for any other 1.3 release, and I would love to see them for 2.0.  I think
the patches for 2.0 will be a lot easier to get into the tree, because a
good portion of them should be localized in APR and not touch Apache at
all.  I am mainly thinking about the 64-bit IRIX stuff.

Ryan

Re: [PATCH] 10x patches for 1.3.9

[Manoj Kasichainula <ma...@io.com>]

On Thu, Dec 16, 1999 at 02:15:46PM -0800, Mike Abbott wrote:
> I think it's really unfortunate that 1.3 is closed to new development

I didn't say there was no chance; I sure don't have the inclination to
veto the patches, because I don't do any maintanence work on 1.3 these
days. And at least a few of your patches were pretty well localized.
But, from what I remember of your patches (and this may have changed
in your latest release), they were all based on some significant
changes for 64-bit Irix support, and some of your patches did touch a
lot of code.

If you want my opinion more specifically, you'll have to wait until
I'm back home and have more than a few minutes at a time to look at
things.

[out of order]
> 2.0 is far from finished.

2.0 is actually not that far away. Once I'm at home again, I don't
think autoconf will take more than a week or two (excluding vacation
time; baby cousins can be very distracting). The other stuff isn't
really difficult; I'm guessing they are one-day projects at most, and
many of them can be delayed until after alpha.

I really think we can release an alpha or beta in January. Bugs will
get shaken out, ports to various platforms will be fixed, and so on.
Depending on how fast development and other contributions go, a
release could happen as early as a few months after that.

> and 2.0 not only is pre-alpha but also has no release schedule at all.

I have only been around for 1.3, but I've only seen release schedules
set a few weeks ahead of time in this group, and they're often
disobeyed anyway. My conception of a release schedule is above;
anybody want to complain or dispute it?

> People who want a high performance Apache server are stuck because I
> can't get my stuff into 1.3 (apparently) and 2.0 is far from
> finished.

I sure won't get in the way of these going into 1.3 (unless we're
talking about 1.3.10).  But, no matter what, They are not stuck no
matter what happens, because they can apply the patches. That's what
the source is for.

I will express my opinion though, which is that 1.3 should be the
stable branch at this point. Cut and paste my comments on the EAPI and
IPv6 patches; they apply equally as well to your patch, except that
(AFAIK) there hasn't been a large user community testing your patches. 

<aside>
It's true that there have been a lot of added features in 1.3 between
.0 and what will be .10, but I don't think this is a good thing.
IMO, when there is a release, that tree should only be updated for bug
fixes and very minor features. Then, we satisfy the desire for new
features by speeding up the release cycle.

Then, we get more people motivated to work on the development tree and
a far less buggy "stable" tree across releases.
</aside>

> By the way, just to end the speculation, would you please vote on
> whether to include my patches into 1.3?  It'd be nice to know for sure
> one way or the other.

Based on what I remember, I'd -1 the patches for 1.3.10, and -0 them
for 1.3.x where x > 10. But, since I haven't seen the patches in a
while, -0 for both.

And there will probably be no more patch transfers from 1.3 to 2.0, so
if you want your work to be available in the 2.0 tree, please submit
patches a decent amount of time before the 2.0 release. Thanks for
your help.

-- 
Manoj Kasichainula - manojk at io dot com - http://www.io.com/~manojk/

Re: [PATCH] 10x patches for 1.3.9

[Manoj Kasichainula <ma...@io.com>]

On Thu, Dec 16, 1999 at 12:15:57PM -0800, Mike Abbott wrote:
> I have ported my patches [...] to Apache/1.3.9

Any chance you'll do this work for 2.0? :) As you've probably seen
from the discussion, large patches on the 1.3 tree will face
significant resistance (including from me). 

-- 
Manoj Kasichainula - manojk at io dot com - http://www.io.com/~manojk/