Re: URN Resolution Paths Considered Harmful

I think that any URN naming scheme that embeds the name of a resolving agent 
for that object's name is a bad idea (IMHO).  Name resolving servers will 
change over time, whereas the intent is for URN names to be unchanging and 
unique for all time.  Unchanging names with embedded resolvers doesn't fit 
well with the fact of changing resolvers...

Three examples from other areas of practice:

1) DNS.  A DNS name (as I have mentioned) does not embed the name of its 
resolver.  This very property has made DNS successful, as changes in the 
resolution path can be made with a minimum of pain.  Admittedly, maintenance 
of the .com domain is messy now, but that is the price of progress.

2) Relational databases.  Having worked for and on CODASYL network databases 
and written relational database applications, I can see why relational 
databases have supplanted CODASYL network databases for most significant 
applications -- flexibility.  If you have hard-wired paths to the data, 
there will always be cases where the path you need does not exist.

3) Domain-based email addresses.  Even for the case of people who receive 
their mail via UUCP, the common practice seems now to be to piggyback on top 
of DNS, such that UUCP path resolution occurs close (in terms of UUCP hops) 
to the UUCP receiving site; the rest of the world does not know or care that 
the physical email address requires an explicit path, as that is hidden from 
the world.  (As an example, an email address of "" which could 
be expanded to "!!jones!root".)  How often, lately, 
have you seen path-based email addresses in journal articles, etc.?  I 
strongly suspect not very often, whereas it was quite common just a few 
years ago.

Computers are very good at keeping track of data and the relationships 
between data.  Let's use computers for what they are good at.
Mark Fisher                            Thomson Consumer Electronics                   Indianapolis, IN

Received on Tuesday, 27 June 1995 08:40:03 UTC