- From: <noah_mendelsohn@us.ibm.com>
- Date: Fri, 21 Nov 2008 20:06:50 -0500
- To: Pat Hayes <phayes@ihmc.us>
- Cc: Alan Ruttenberg <alanruttenberg@gmail.com>, Harry Halpin <hhalpin@ibiblio.org>, Jonathan Rees <jar@creativecommons.org>, "public-awwsw@w3.org" <public-awwsw@w3.org>
Reflecting on the note I sent below, I realize that there's a subtlety I didn't deal with. There are at least two abstractions we might want to identify: 1) A representation in the sense of a particular set of bits sent at a particular time from an HTTP resource to a client or proxy, or conversely, such a representation sent upstream with PUT or POST. By this definition, any representation involved in a subsequent interaction would necssarily be a different one. 2) A representation defined to be the information sent with a GET, PUT, POST, etc. By this definition, it's at least possible that the same URI would be used to identify the representation that resulted from, say, two or more GETs. I believe the natural way to go down this path is to allow the same URI to be used to identify representations with the same content, even if provided as representations from very different resources. For example, the representation that is Content-type: text/plain with entity body "hello world" could have the same URI even if returned on GETs to many different resources. When I claimed that I wanted to indicate that a given representation was poorly formed, I could have meant in the sense of (1) or (2). If I want to make a statement that a particular representation was received at 4PM, I probably mean in sense (2). Or maybe for that case I need a URI for the HTTP response, which contains a representation but potentially contains additional information as well. Anyway, I realized that there is potentially such an ambiguity, and I think we should be very clear which of the possible abstractions we refer to when we decide to identify a "representation" with a URI. Noah -------------------------------------- Noah Mendelsohn IBM Corporation One Rogers Street Cambridge, MA 02142 1-617-693-4036 -------------------------------------- Noah Mendelsohn 11/21/2008 06:22 PM To: Pat Hayes <phayes@ihmc.us> cc: Alan Ruttenberg <alanruttenberg@gmail.com>, Harry Halpin <hhalpin@ibiblio.org>, Jonathan Rees <jar@creativecommons.org>, "public-awwsw@w3.org" <public-awwsw@w3.org> Subject: Re: statements about resources vs. representations Pat Hayes writes: > I think what Harry should have said is that they are too > ephemeral for someone to want to give them an enduring name or > identifier. But there are other ways to refer to things than > baptizing them with a URI for all time. On this I don't think I agree. We're talking about the Web here, and what's more, I think a representation is an information resource. I mean, not only can the thing be represented as a computer message, the whole point of it is to be sent in a computer message! The key architectural imperative for the Web is "Identify with URIs." I see no reason why, in cases where you do want some means of identifying a particular representation, a URI wouldn't be the way to do it. When I make that choice, I get a variety of advantages: I can make Semantic Web statements about the representation (it was buggy, it took a long time to arrive, it was cached at proxy p1, etc.) in the natural way without resorting to indirection; I think I could even choose to run a server that would respond to GETs with representations of, well, the representation. I think the usual rules of the Web apply well here: when you need to identify something, do it with URIs. Noah [1] http://www.w3.org/TR/webarch/#pr-use-uris -------------------------------------- Noah Mendelsohn IBM Corporation One Rogers Street Cambridge, MA 02142 1-617-693-4036 --------------------------------------
Received on Saturday, 22 November 2008 01:07:35 UTC