From what I've seen & tested so far, +1 to this patch. And rolling it into Graham's.
On Sunday, March 4, 2001, at 11:49 AM, Christian von Roques wrote: > Chuck Murcko <[EMAIL PROTECTED]> writes: > > http://dev.apache.org/dist/patches/apply_to_1.3.19/ > > In the mean time I've made some small whitespace and comment fixes to > my changes to Graham's HTTP/1.1 patch. Below you'll find my changes > enabling Apaches mod_proxy with Graham's patches applied to make use > of HTTP/1.1 persistent connections with its clients. > > The patch changes mod_proxy to write the reply-headers using > ap_send_http_header() instead of directly using ap_bvputs(). This not > only simplifies mod_proxy, in my opinion at least, but enables it to > make use of the features of Apache's normal header and persistent > connection machinery. The patch is relative to apache 1.3.20dev from > CVS with Graham's HTTP/1.1 patch from > http://dev.apache.org/dist/patches/apply_to_1.3.19/ applied. > > I've been swamped with other work and didn't find as much time as I > would have liked to test the patch. > > I've found one peculiarity so, IE5.5 makes good use of persistent > connections when directly accessing some pages via HTTPS, but this > doesn't work when IE uses CONNECTs through a Squid to access the HTTPS > proxy. The connections stay open, but are closed as soon as IE does > the next request. This even is the case when IE's optiona setting to > use HTTP/1.1 persistent connections through a proxy is enabled. > Because Squid has to be transparent for HTTPS to work through > CONNECTs, it probably is a problem with IE and not the patched > mod_proxy. Is this a known issue with IE? > As for the stuff Christian talks about above, the CONNECT method definition is nothing more than a proposed RFC (did it ever get submitted?). This behavior sounds consistent with the original CONNECT definition though, and likely is not a problem with IE. I believe we should see the same behavior through httpd's CONNECT proxy as through Squid's. CONNECT defines a dumb, nocache proxy, IIRC. I'll see if I can dig up the original definition. Chuck Murcko Topsail Group http://www.topsail.org/
