- From: Jim Davis <jdavis@parc.xerox.com>
- Date: Wed, 1 Jul 1998 14:22:09 PDT
- To: <w3c-dist-auth@w3.org>
At 01:30 PM 7/1/98 PDT, Babich, Alan wrote: >(1) GET on a reference returns 302 as per John's proposal. >(2) PUT works on a reference, unless what is being put >is not a reference, in which case it fails. Sorry, I'm confused. What do you mean? You can't PUT a reference. You can create one with MKREF, but not with PUT. In John's semantics, a client that does a PUT to a referential resource gets back a 302 and can then decide what to do. (Either report an error to the user or retry the PUT to the new URI.) If the user desires to replace the referential resource with one of some other type (either 'plain' or even a collection) the client first does a DELETE then a PUT (or MKCOL). This definition allows suitable clients to make ref resources act EXACTLY like plain resources, as far as the user is concerned, if the user wishes. That's real flexibility. And it also allows clients to treat them differently, if desired. Look at it this way. With most current servers you can create a redirecting resource by writings a suitable config file somewhere. So this capability already exists in the Web. The only difference is that now it will be possible to create such redirecting resources from the client side alone, by invoking HTTP methods, without creating or writing any server-side config file. It's a winner, I think.
Received on Wednesday, 1 July 1998 17:21:47 UTC