Re: HTML icon set was: Additions to the CGI archive

Dave_Raggett (dsr@hplb.hpl.hp.com)
Fri, 7 Jan 94 18:29:44 GMT


Roy T. Fielding writes:

>> I have already suggested that we go ahead and start defining URNs
>> based on the same domain syntax as host names, see my message of the
>> 4th January.

> I don't recall seeing a message like that. Could you send me (or www-talk)
> another copy? I've noticed that the listserver has been acting strange
> lately -- one of Rob McCool's messages spent two weeks in Switzerland
> before I received it here. Come to think of it, Rob's message arrived
> on the 4th...hmmmmmmmmm Was that the day they changed to www0.cern.ch ?

well here it is again:

Maybe we ought to come up with a de facto URN mechanism, e.g.

URN ::= DOMAINS [; SELECTOR]*
DOMAINS ::= DOMAIN . DOMAINS

The DOMAIN is an opaque alphanumeric string which is used
to track down the domain servers a la DNS and identify the
individual URN. (it doesn't have to be meaningless though).

In this model, URN's can stand for single objects or for
families of related objects, which can be selected from
according to the value of the SELECTORS, e.g. language,
most recent version, ...

The binding process converts such URNs into one or more
URLs plus info differentiating family members and giving
other useful info, e.g. ownership, date created, ...

While people are developing suitable binding protocols and s/w
we can start using URNs by introducing a URN attribute as
previously described.

Dave