At 03:16 AM 6/18/2010, Paul Ruizendaal wrote: >The CGI protocol specifies that all http headers of the request >are passed on to the cgi script as environment variables, using the >following rules.... >.... the header value is left unchanged >So for a definition of HTTP_HOST you have to look at the spec for the >Host: header of the http protocol.... >And yes, Richard is correct: the port number is obligatory in the >host header if different from "80".
Thanks, that connects the dots now. Although I can see how it could be read otherwise, I wasn't actually doubting Richard. I was just surprised that a definitive answer wasn't immediately discoverable so that I could cite an authority if I wanted to try to push a bug report or patch back at the thttpd folks. >.... >> There are also SERVER_NAME and SERVER_PORT. Thttpd does supply >> those, and they are named explicitly in every CGI document a >> quick search turned up just now. Perhaps the host name should be >> assembled from those two, omitting the port if it is 80 of course. > >This won't work: SERVER_NAME is not necessarily the same as the >host name of the request.... I knew after I posted it that that was too obvious to be right. Keeping track of this kind of thing is why I went back to embedded systems and stopped trying to keep up with web app development. That and JavaScript. ;-) >... >However it is easy to patch thttpd to use the SERVER_PORT data to >fix up the HTTP_HOST information (same libthttpd.c file): Thanks for the pointer to the right spot to patch. I had to tweak it a bit because build_env() isn't variadic despite its printf()-like appearance, but I got something that works for me. All in all, this is an eminently practical answer in my case. I'm certainly not going to do virtual hosting on a NAS box, and this small fix enables thttpd to host fossil and work as I hoped. Thanks again. Now I just need to automate web page with a usable directory listing of repositories, and a cron job to synch them once in a while... Ross Berteig [email protected] Cheshire Engineering Corp. http://www.CheshireEng.com/ _______________________________________________ fossil-users mailing list [email protected] http://lists.fossil-scm.org:8080/cgi-bin/mailman/listinfo/fossil-users
