Rob, this is a clear presentation of the issues. I have
a few nits and a few strategic questions.
Your definition of an Instance doesn't quite allow it to be the
result of a query that might have differing results (assuming
the server knows where I am, I could say, "send me the
addresses of all the bookstores in my area that carry ORA books").
I think "indistinguishable" is unneeded: the owner says they're
the same (for the purposes at hand) and that's good enough.
You say that "the actual content of a URN is that to which it
refers," which doesn't make a lot of sense when taken literally.
What's your thought here?
Your concluding discussion of intellectual property, while
relevant to Internet publishing and URNs, doesn't seem necessary in
connection with Registrar, which is a concrete proposal for
making URNs work.
URNs may never be destroyed, yet I might want to give URNs to
every chunk of every section of every book in every series we publish,
for all editions. (Is that right?) So a whole lot of URNs are
going to build up, and quickly, too. How would you figure out
when to delist a URN? Wouldn't the number of URNs eventually
make it slower to use Registrar than to try some URLs in the
first place? This is a general issue with URNs, not specific
The amount of info you require for the URN Record Format is
considerable. Is it all needed? On a related point, why should
Registrar support querying for CREATOR? If you are an individual
who just wants to put some stuff online, you are likely not to
bother with registering your URN if it's this much work. Is that
a bug or a feature?
-- Terry Allen (firstname.lastname@example.org) Editor, Digital Media Group O'Reilly & Associates, Inc. Sebastopol, Calif., 95472