* Tim Bray <[EMAIL PROTECTED]> [2005-08-13 16:50]: > The fact that I use HTTP URIs for identifiers reflects my > belief that good Web citizenship requires that once something > is published and its URI widely disseminated, it should never > ever be moved; so in my case this scenario is unlikely to > arise.
I don’t know if you can reflect any belief by your choice of scheme for your atom:id values, though. As far as the consumer is concerned, the string is opaque and meaningless anyway, so whatever you put in there matters naught as long as it’s a URI. Therefore, *I* believe that authors of database-based CMS software of any form should be advised to use UUIDs, because a) it is instantly obvious that UUIDs have to be stored with their respective entry, so there’s no temptation to derive the ID dynamically from any other part of the data, and b) UUIDs are fixed-size and the binary form is very compact, which makes them extra easy to deal with in a database. Regards, -- Aristotle Pagaltzis // <http://plasmasturm.org/>
