- From: Sandro Hawke <sandro@w3.org>
- Date: Thu, 02 Jan 2003 05:59:06 -0500
- To: "Roy T. Fielding" <fielding@apache.org>
- cc: www-tag@w3.org
> > terms. The real distinction is just whether you focus on the > > collection of information from which the web server generates its > > response or focus past the server, as Roy does, on the subject of that > > information. > > No, the real distinction is whether or not you can give a URI to > something that doesn't exist yet. You cannot do so if what you are > naming is a document, whereas you can for a concept that is only bound > to a representation at the time of a request. Abstraction #33 says nothing about documents. If I rephrased it to use the word "document", it would be in the form of "living document" or "maintainable document". A lot of people, including you (it seems) think of a document as a fixed artifact. TimBL does not; he seems to think of a document as a mutable thing. One can exist unwritten, then in draft form, then rewritten over the years. It's common to talk about the different drafts and published revisions of a book as the same book, so I see his point. This ambiguity makes me uncomfortable, though, so I try to avoid the word "document". I said: Abstraction 33: An http URI is most strongly associated with a repository or collection of information. When you GET that URI, the bytes returned convey that information. I never said the repository had be somehow materialized or available at the point of name-assignment. The bytes sent back to a client during a successful HTTP GET are a serialization (in some language like HTML+English, JPEG, or RDF/XML) of the information in the repository at that time. These repositories might also be called databases, database instances, knowledge bases, files, data structures, objects (in an OOP system), etc. To a web designer they might be called web sites, or web pages. (Many of these terms, especially "web page", suffer from the same sense of immutability as "document"; people sometimes say a "I got a different page" when the content has changed. Even the word "database" has the same ambiguity, although I think "DBMS instance" or "database table" might be okay.) For me the clearest mental image is of a (virtual) location where information might or might not be present; like a bulletin board, or whiteboard, or a desk covered with papers. The information present at the location may change over time, but the location itself persists. I sometimes call this the distinction between the "content" and the "container". URIs are attached (in abstraction 33) to a container. (And the web client does not visit the location; it talks to a server agent who is or pretends to be at the location. The server observes the information-content at the location and (when things go well) describes it to the client. Sometimes the server modifies the information content on behalf of the client.) It is these locations which search engines index, based on the content available at the time of traversal. "Cool URLs dont change" is about keeping the container thematically consistent, so when people return to a place, or arrive there on someone's recommendation, they get roughly the intended experience. So yes, the contents of the repository may change over time and may be indexed and linked-to before being actually made available. I don't have a proper write-up of this model, but I do have two currently-idle slightly-out-of-date drafts which talk about it in more detail, for people who would like a more detailed exposition [1] [2]. > Use the technology and think about what it implies. This is not just > my opinion, it is my experience, and it isn't going to change. The question here is what mental model of the web (if any) the TAG should recommend that people use. Experience can help you develop and test a mental model, but it is does directly constitute one. Many educated and intelligent people used to believe they experienced the sun moving around the earth and heavier things falling faster than light ones. And yes, they put Galileo on trial for heresy rather than changing their minds, but that doesn't mean they were right. Some of us, in trying to develop the semantic web, have observed things about the web which make the mental model you suggest seem not as accurate as it once did. I understand if you're too busy to think about new alternatives right now, but please don't claim they are inferior just because your model has served you well enough so far. (One observation you can reproduce by writing various formal assertions about the URI "http://www.w3.org/Consortium/". It turns out that sometimes (as in EARL) you want to talk about some marked-up text (the content), sometimes (as in ACLs) you want to talk about a maintainable information-repository (the container) and sometimes (as in the membership database) you want to talk about a consortium (the subject). If you say the URI is directly attached to the subject (as in abstraction 102) and recognize that other URIs may well have the same subject, it becomes impossible to assert things about the container and its the contents. If you use abstraction 33, you can talk about all three.) -- sandro [1] http://www.w3.org/2002/11/bees-and-ants/paper [2] http://www.w3.org/2002/11/webarch4/v3
Received on Thursday, 2 January 2003 06:03:21 UTC