On Tuesday 27 November 2001 04:59, Geoffrey Talvola wrote: > > Of course it should come with a big notice saying that it's for > > development and testing only. It's been lightly tested with HTTP > > GET requests and seems to perform just as well as Apache with > > some simple Cheetah servlets (180 req/sec for 600 req with a > > concurency of 100). > > Well, that was the case with the HTTP server that used to be > included with Webware. It was marked as experimental, but it was > fast and seemed to work OK with simple requests. Unfortuately > people kept on reporting problems that happened as they got a > little further into using it under varying conditions, and nobody > had the time to fix it up. > > Unless someone's willing to commit to real testing and fixing bugs, > I'd say let's not include it in Webware and commit the same > mistake.
Or we could include it with a RTFM policy on help requests. The old one was way more complicated that it needed to be because of the Async architecture, which made it a nightmare to understand and debug. It was also affected by the instability of the Async code as a whole. If we are going to include a bare-bones webserver with Webware, I think this is the route to take rather than trying to hack Medussa/ZServer/Twisted or bundle a third-party webserver like Apache or Xitami. Medussa et al. are not a simple beasts to grok. Bundling a third-party server would probably lead to a slew of help requests related to it rather than WebKit. Cheers, Tavis _______________________________________________ Webware-discuss mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/webware-discuss
