- From: Walden Mathews <waldenm@optonline.net>
- Date: Mon, 03 Feb 2003 09:44:40 -0500
- To: "Roy T. Fielding" <fielding@apache.org>
- Cc: www-archive@w3.org
----- Original Message ----- From: "Roy T. Fielding" <fielding@apache.org> To: "Walden Mathews" <waldenm@optonline.net> Cc: <www-archive@w3.org> Sent: Monday, February 03, 2003 2:52 AM Subject: Re: Valid representations, canonical representations, and what the SW needs from the Web... > > Resources which are representations seem to be among the > > most confusing critters in the REST world. If you have a resource > > that is "the HTML document that describes X", may it be > > identical with its representation dished up on GET, or is the > > bunch of bits that represent "the HTML document..." somehow > > still a shade different from the resource in this case. Kind of > > like 1/8th of a dimension away? More like an infinite number > > of levels of abstraction.(?) > > It is always different in the sense that one is a resource > (identified and accessible) and the other is a result of an action. So when a client requests (using GET) "the HTML document that describes X", the representation returned is not that document, but merely a string of bits that represents it? Yet it seems at other times it's been claimed that when the resource is something concrete, say ... ME, then the representation returned may be a document. So a document can be a representation, unless the backing resource is itself a document? You may have guessed that I'm trying to figure out whether one end or the other of this model is nailed down to something concrete. I'd be tempted to theorize that "representation" is always a "string of bits", except that REST applied outside of Von Neuman computing wouldn't have representations that fit that description either. > Anything that can be identified can be a resource. It follows that > you can identify things with very specific constraints on > representation, > such as a specific version of a document, and give them a URI. One thing I'd like you to clarify is that when you say "identity" or "sameness", are you implying a totality of sameness, so that between A and B (two names for something that may be one thing), that if A is identical to B, then there can be no view from which a distinction can be found? Or is identity contextual in the sense that if in the current view(s) there are no distinctions, then identity is operable? Either way, it seems as though identities are always going to be vulnerable to disintegration upon the discovery of new information or views, but I guess that's okay. Thanks, Walden
Received on Monday, 3 February 2003 09:44:49 UTC