Apache Developers,

Here is the second patch in my series that increase Apache's performance
ten-fold.  This one streamlines request parsing.

The vanilla Apache parsing functions read_request_line() and
get_mime_headers() make multiple passes and many copies.  I wrote new
functions to parse common HTTP request lines and headers in a single
pass, copying minimal data.  The streamlined parsing routines
quickly_read_request_line() and quickly_get_mime_headers() do not handle
all the edge cases, nor do they have to:  When they detect a construct
they cannot parse quickly they give up and fall back on the vanilla
Apache parsing functions.  They are meant to parse the simple, common
cases quickly and to recognize when they cannot do so.

I also included a couple minor tweaks that speed up the vanilla parsing
path.

This patch may break binary compatibility of modules because it adds
fields to the request_rec structure.

The base for this patch is Apache/1.3.6 + 10x performance increase patch
#1.

Again, I welcome your review, questions, and feedback.
--
Michael J. Abbott    mja@sgi.com