httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
Subject Re: cvs commit: httpd-2.0/support ab.c
Date Tue, 03 Apr 2001 13:18:32 GMT

If you guys are focused on AB.C again you might
still want to consider adding full IETF Content-decoding
and chunked transfer capability. 

Already done and been around for almost a year now.

It was offered up to the group some time back but there
was no reaction and we were told to just 'put it somewhere
where people can find it' so that's what we did.

You are more than welcome to just grab this 
version of ApacheBench that supports full IETF Content-decoding,
and 'Transfer-encoding: chunked' and do whatever you
want with it. The only license on it is still YOURS. It's
just sitting there waiting for you.

The 'Home' page for this 'enhanced' version of ApacheBench
is here...

Pre-compiled binaries and link to full source code are on that page.

or... direct link to an HTML page that will display the full source 
in your browser window is...

There is still only 1 source code file... AB.C.

The top of the code explains exactly how to add Content-decoding
support to ApacheBench, what the issues are, and how to simply
add 'ZLIB' to it which is what it uses to decompress the actual
'Content-encoding: gzip' responses it gets so it can show 'Virtual'
HTML byte throughput rates as well as 'RAW' compressed data
transfer rates. You do NOT have to actually 'decompress' anything
unless you want these Virtual throughput statistics.

Everything to do with the 'enhancements' is encased in the 
following 'define' in source code...


You should be able to just 'cut and paste' those sections
into ANY existing copy of ApacheBench to add the new feature(s).


This 'enhanced' version of ApacheBench was written against
Apache 2.0 and will be looking for 2.0 headers. The base code
for it was taken from the first APR enabled version of ApacheBench
that appeared on the horizon some time back when.

There is nothing in the AB_CAN_DECODE sections that 
REQUIRES Apache 2.0... but the base AB.C itself was 2.0 version.

Based on the email we get it would seem that thousands of people 
seem to be using this puppy but only a few have reported any
problems. The one problem that does seem to re-surface a lot
is people saying that this particular cut of the code has some
problem running ZLIB and maintaining concurrency when the
numbers are really high like -n 10000 with -c 100 or something.
Has something to do with ZLIB not being able to 'keep up', methinks,
but in all honesty I have not had the time to go back and verify
whether the problem is ZLIB or not. We usually don't use the 
'-c' option much and most others don't either and without any
high concurrency factors it all works like a champ.

Kevin Kiley

View raw message