- From: John C. Mallery <jcma@ai.mit.edu>
- Date: Thu, 15 Aug 1996 17:57:53 -0400
- To: "Roy T. Fielding" <fielding@liege.ICS.UCI.EDU>
- Cc: http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com, bug-cl-http@life.ai.mit.edu
At 1:29 PM -0700 1996-08-15, Roy T. Fielding wrote: >Before people start getting jumpy ... > >The Apache Group is not planning on releasing any server that uses >the native HTTP-version "HTTP/1.1" until after the IESG approves the >draft for RFC status (i.e., fixes that version in stone). Since all >of the HTTP/1.1 features can be tested as an HTTP/1.0 server (assumimg >you are careful not to send HTTP/1.1 features to HTTP/1.0 clients >where it is noted in the spec that you MUST not do so), I strongly >suggest, for the sake of interoperability, that no application be shipped >with the "HTTP/1.1" label until after we all know that there will be >no further requirements placed on HTTP/1.1 applications. > Are you saying that 1.0 clients do not ignore headers they don't understand? Or are you trying to say don't send chunked encoded stuff to 1.0 clients? I checked every occurence of MUST in the spec to make sure we were conforming, and I believe we are. CL-HTTP responds to 1.0 clients with 1.0 responses. 1.1 responses are reserved for clients or proxies that also advertise 1.1 functionality. Most of my user base is behind firewalls, with one prominent exception, and they like keeping their research on the cutting edge. Of course, since we are in lisp, upgrades only take a couple of minutes. Servers don't go down. Just load patches. No big deal.
Received on Thursday, 15 August 1996 15:00:47 UTC