Justin Erenkrantz wrote: >On Tue, Jan 08, 2002 at 04:20:40PM -0800, Brian Pane wrote: > >>Don't we already know that it's not beta-quality, based on the >>problems with 2.0.30 on daedalus? >> > >I thought that the problems on daedalus were related to CGIs >reading one char at a time? Wouldn't a simple change to make >the CGIs use buffering (at the APR-level) solve this? -- justin >
The single-byte reads for CGI headers definitely are bad, but they were in 2.0.28 too, and that didn't have the problems that .30 does. So the root cause is probably something else. Regarding a fix for the single-byte reads, I agree that we should add buffering. I posted some notes on fixing this in the httpd code last Friday (subject: "CGI single-byte reads Re: 2.0.30-dev load spiking [was: upgrade to FreeBSD 4.5-PRERELEASE]"), although changing APR to allow buffering on pipes might work just as well. --Brian
