[I've kept this on the list because this, in spite of the topic drift, 
it still seems relevent to the design discussion]

Well, you're probably right.  There needs to be a very careful balance 
between too complex and too simple.  The URL is bloody simple and
therefore everyone and their brother can hack together some HTML with
URLs and volia, they are a web publisher.  This has helped the web
grow...

I still can't get the RFC, so, this is from my past (faulty and
failing memory) but the URN concept requires a bit of support
structure.  You need "official" sources for your URNs and you need URN
servers, etc.  That works well for DNS, but, you end up needing quite
a mgmt structure to support it.  With URLs you, the publisher are the
offical source.

I had the glimmers of a clever solution here, but, I've spent the day
with C++ and the brain is suffering.  The problem comes up that you
don't know who's linked to you, and, therefore if you change the
location of anything you have no way of notifying anyone.  Once you
have a page that you've linked into anywhere in the web you've made
that URL public.  You can keep everything forever and use redirects to
keep the outside links working as you update your internal structure,
but, boy is that messy.

Oh well.

cheers

bruce


Mark Guzdial writes:
 > >Lex Spoon writes:
 > > >
 > > >     2) Links are so easily broken.  Maybe *all* links should go
 > > > through a level of indirection on a stable server?  Do links *really*
 > > > need to be all that human readable?
 > > >
 > >
 > >Actually, this was the origional idea.  URLs were a quick hack until
 > >URNs worked.  URNs were going to work a bit like DNS does with a
 > >distributed database so you would not have broken links and give you
 > >that extra level of indirection.  Hmm, this was '94 or so :-)
 > 
 > On the other hand, I've read some analysis (armchair quarterbacks, I admit
 > :-) that suggest that the Web succeeded where previous hypertext efforts
 > failed (e.g., Intermedia, Xanadu) explicitly because URL's were
 > one-direction only.  All previous efforts made sure that links couldn't
 > break, and this analysis claims that that extra bit of engineering is what
 > pushed the system over the edge to overly-large and less-adoptable.
 > 
 > Mark
 > 
 > --------------------------
 > Mark Guzdial : Georgia Tech : College of Computing : Atlanta, GA 30332-0280
 > (404) 894-5618 : Fax (404) 894-0673 : guzdial at cc.gatech.edu
 > http://www.cc.gatech.edu/gvu/people/Faculty/Mark.Guzdial.html
 > 
 >