Re: [xml-dev] Best Practice for URI construction?

[Peter Hunsberger <peter.hunsberger@gmail.com>]

On 12/12/05, Michael Kay <mike@saxonica.com> wrote:
> >
> > I was giving examples of possible results, for any given
> > implementation it's should never be ambiguous.  Eg; without the query
> > parameters it identifies the search screen (perhaps). With the query
> > parameters it identifies a single patient (assuming one exists for the
> > given parameters). An alternate implementation might be that without
> > query parameters the results is a list.
> >
> > Or where you worried about something else?
>
> Yes, I'm worried that we're trying to solve the meaning of life the universe
> and everything when we clearly lack the intellectual apparatus even to
> define the problem.

Given that it's tempting to think that URIs can somehow identify
everything, and that people tend to think that classes etc. can
somehow model everything I can see how you might come to that
conclusions.  However, no, that was not my intention.  I was just
trying to add another perspective on the issue of what to put in the
query string vs. what to put in the rest of the URI.

>
> You're right that if we stick to "implementations" the problems are
> tractable. I wouldn't say "implementations", I would say "closed systems".
> When we do information modelling we ask questions like "what is a flight?",
> "if a flight involves a stopover, is that one flight, two flights, or
> three?", "if an extra plane is laid on to handle extra demand, is that the
> same flight or a different flight?". I know how to tackle these questions
> within the confines of a closed system where we can agree the terms and what
> we mean by them. A smallish group of people can get together and decide on
> precise definitions of the terms they are using within a limited domain of
> discourse.

Interesting example since I once had to design and build a reservation
system for a company that managed airline reservations across multiple
business partners.  That aside, if the issue is creating URIs, then it
seems to me you're usually working on a local or closed system?  
Personally, I certainly don't expect to be able to reuse a URI from
Northwest Airlines on a Delta Web site....

> I simply don't believe that it can be done universally, and what worries me
> is that there seem to be people who think it can. What I mean by "flight"
> depends on the conversation I am having at the time, and calling it
> http://www.saxonica.com/vocabulary/flight instead isn't going to change
> that. OK, we could define 120 different URIs to cover the different precise
> meanings of the word, but that would only reduce our ability to communicate
> with each other. There's a good reason why language is fuzzy and full of
> nuance: if it were possible to develop a precise and unambiguous and
> unchanging vocabulary we would have evolved one years ago. Deciding that
> every distinct concept is going to have a distinct URI is just simplistic:
> like tons of bricks or piles of sand, concepts are amorphous and lack clear
> identity. Should we talk patents?

Sure, but I don't see how any of that relates to my proposal?  If
anything it's the exact opposite: if you haven't already got a good
domain model for a particular concept (ie; if it isn't already well
defined within your business), then that particular concept may not be
a good candidate for URI construction within your business...

--
Peter Hunsberger