httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
Subject Re: ab Pt 2
Date Sun, 15 Oct 2000 12:23:59 GMT

In a message dated 00-10-15 09:40:11 EDT, writes...

> There's a note in ab that strstr is a performance drain.  The diff 
> attached tries to remove strstr from a few places and changes the way 
> that we check things.  It works and is as fast, if not slightly faster, 
> that the current code but I'm sure it can be improved on.

Just about every robot or browser code you will look at is using
s = strstr(c->cbuff, "\r\n\r\n");
to check for 'end of header' in the first buffer received.
It's SOP.
If one of the goals of ApacheBench is to 'imitate a browser'
and get true real-world timing stats then leaving these kinds
of things in it is actually desireable.

BTW: I submitted a complete re-write of ApacheBench that
addresses a lot of these issues and also addes full IETF
Content-decoding support and full Transfer-encoding: chunked

Has anyone looked at the submission?
Did it get lost in the shuffle?
Do I need to send it again?

Complete source code for Enhanced ApacheBench is available
online directly from the following URL...

Kevin Kiley
CTO, Remote Communications, Inc. - Free IETF Encoding Server - Free Enhanced ApacheBench - Free Apache Content 
Acceleration module

View raw message