Sunday, July 13 2003

Are you out of your mind? Shelby asked. There's already so much food that no one will have room!

That's what they all say, I replied. And then it arrives and, with it, The Silence.

Philemon nodded knowingly.

Editor's note: it was pointed out to me, today, that I was clearly very drunk by the time the following occurred.

The problem is, I told the Princess, that I've always been very drunk by the time I get around to making these. So, I never remember what I did the last time or if I'm actually doing it right.

Both the Princess and Frances licked their plates clean.

This is not an N3 problem. It is an interface problem and since every single Semantic Weenie I've ever met invariably ends up saying people aren't going to write this stuff, machines are it's a very real problem. If, in the context of writing recipes, we're going to say that people can just write and the computer will figure out how to translate an ingredient in to a URI then I, as the code-monkey, suddenly have to become an arm chair linguist, keep up with every imaginable derivation of a word and maintain a potentially enormous lookup table in memory every time the user does, well, anything. I happen to think that having something like URIs for things like ingredients is a good thing. But you are out of your fucking mind if you think anyone is going to write URIs or some specialized subset of a language that is URI-friendly. This segs nicely into the whole URI as identifier thing in the first place. I am told that there is no requirement that a URI actually be resolvable; that it might be is only a courtesy. Prompting me to remark, during a different conversation :

Okay, so we've gone from well-formed trees to loosey-goosey graphs and from URIs actually pointing to something to simply being really long phrases with a bunch of arbitrary and meaningless cruft tacked on at the beginning. Got it.

I can appreciate that if you don't think that URIs need to be resolvable (let me just ask: what streak of madness and slapstick cleverness inspired you to choose URIs as the basis for your unique identifiers?!) you are not constrained by the sorts of naming conventions we out here in, well, URI-land live with every day :

Are for generating the former from the latter (URI -> title) I'm willing to investigate it, but I fear that it will fall prey to various character encoding and filesystem-related naming "issues", particularly if the resources at the end of each URI actually exist.

(I preemptively slap, upside the head, the second greasy propeller head who suggests I simply maintain a very complicated set up rewrites for mapping all this stuff back to machine readable filenames.)

The Semantic Web is doomed if I, as tool-maker, have to throw an exception every time your grandmother writes something with two adjancent dots or eggs instead of egg.