I disagree. I think stuff like Mosaic shows that "If you build it,
they will come." Link has not been used because no software has been
deployed to exploit it. There are pretty clear applications for this
mechanism, but nobody has needed it badly enough to code it up. Or if
they did, they weren't happy enough with their stuff to write a
complete proposal and spread it around. Or they didn't have time to
code up a nice user interface to the functionality (which is critical,
really).
It is not the business of an IETF group like this to design this
thing. It's our business to review proposals that are already pretty
well baked.
I think another reason why LINK hasn't been widely deployed is that
the links between documents are rarely "static" or "global" enough to
just stick them in HTML markup and be done with it. Consider the
"print the document tree rooted at XXX" application: Some folks will
want the writer's notes and annotations, and some won't. A document
might be shared among several corpuses, and the children of that
document might be different depending on the corpus you're interested
in.
Have a look at LINCKS for an architecture that deals with this problem
_very_ well. It effectively puts the job of aggregating documents --
or breaking them up -- on the server side, rather than the client.
My notes:
http://www.w3.org/hypertext/WWW/People/Connolly/technologies/lincks.html
Lincks-to-HTTP gateway:
http://www-lincks.ida.liu.se:8080/w3l/DB/httpd/
"LINCKS --- A Platform for Computer Supported Cooperative Work"
http://www-lincks.ida.liu.se:8080/w3l/DB/httpd/TS:1/TP:6-document%20overview
Dan