- From: Roy T. Fielding <fielding@gbiv.com>
- Date: Thu, 13 Mar 2008 17:23:18 -0700
- To: Henrik Nordstrom <henrik@henriknordstrom.net>
- Cc: Julian Reschke <julian.reschke@gmx.de>, John Kemp <john@jkemp.net>, HTTP Working Group <ietf-http-wg@w3.org>
On Mar 13, 2008, at 5:07 PM, Henrik Nordstrom wrote: > On Mon, 2008-03-03 at 17:24 +0100, Julian Reschke wrote: >> I think that would relax it too much; it's almost making it useless. >> >> My preference would be to use Mark's text, and also relax the "SHOULD >> believe" requirement for the client. > > What exacly is the problem with the client requirement "SHOULD be > followed"? It is not an interoperability requirement and there are no known implementations. > It's a recommendation that IF the server has indicated a list of > supported methods the client should not try to use methods outside > this > set unless is has specific reasons to go against the servers > recommendation. In other words, it says nothing useful. There are no interchangeable methods in HTTP. Allow only lists the methods that are known to be allowed, not the set of all methods that might be allowed. As such, "SHOULD be followed" is wrong. Allow is useful for one and only one purpose: to discover those methods that the server is willing to advertise, usually in response to OPTIONS. ....Roy
Received on Friday, 14 March 2008 00:23:57 UTC