- From: Jim Gettys <jg@pa.dec.com>
- Date: Wed, 8 Sep 1999 08:13:21 -0700
- To: "Josh Cohen (Exchange)" <joshco@exchange.microsoft.com>
- Cc: "Roy T. Fielding" <fielding@kiwi.ICS.UCI.EDU>, http-wg@hplb.hpl.hp.com
Here's the story, folks (short history lesson): The lack of host information meant that IP addresses has had to be assigned for all virtual host web sites. Not only did this waste IP addresses (getting to be a scarce resource), it caused MAJOR headaches among major ISP's doing Web hosting (think about the problems caused by adding the 257'th host to a web server on a class C net (or whatever you call the CIDR equivalent these days). So certain (more than one) IESG/IAB members insisted this get fixed in such a way that it COULD NOT UNDER ANY CIRCUMSTANCES be gotten wrong in all software going forward. There was no confidence in people getting it right otherwise. At least one of these people said 'I don't care if HTTP/1.1 has anything else', as a way to show the vehemence of the problem. Ergo, the requirement to report buggy clients with an error. And note Roy's remark about proxies... Might there be some further optimization possible? Maybe. But we sweat blood over them words, as I remember. And they were written to enable transitioning to absolute URI's in a V1.2, as I also remember (we had a problem that would have made that hard at one point, that did get fixed). If you insist on riding this horse, you should start by refreshing your mind with all the discussion on the topic, and of exactly how the spec words got to where they are. The issues were more subtle than they first appeared, as I remember the discussion (and I don't remember all the subtleties). - Jim
Received on Friday, 10 September 1999 01:33:57 UTC