Re: REL and REV attributes (Was: More comments on HTML 3.0)

Paul Burchard (
Wed, 26 Apr 95 16:01:23 EDT (Craig Hubley) writes:
> [Paul Burchard <> writes:]
> > Drop REV, but add general relational operators to the syntax
> > of REL. E.g., REL="Reverse(Child)" in place of REV="Parent",
> > and REL="Join(Glossary,Author)" to refer to the author of the
> Are you classifying the link, or what one finds at the other end ?
> I had understood that REL/REV were primarily to define the
> relationship.

Yes, that's correct. To be precise: REL declares that the ordered
pair (link-source, link-target) satisfies a certain binary
relation. What I'm suggesting is that it would therefore be natural
to provide relational operators as well.

> In what sense can Document B be the
> 'Join(Glossary,Author)' of Document A?

Well, "Document B" would typically be a human document. :-) E.g.:

Glossary by <A REL="Join(Glossary,Author)"
HREF="">John Doe</A>

> Meanwhile, bear in mind that 25 years of hypertext
> research has proven the value of naming links with simple
> strings determined by the author, with NO semantics
> assumed by the basic (non-application-specific)
> software. Building a formal type system on the strings is
> not necessarily as widely accepted a practice or skill.
> In hypertext systems that tried to do this, like Xerox
> NoteCards, users usually fell back to 'ad hoc word
> association' anyway.

Any good references? The problem is that ad-hoc RELs with no
semantics are what we have now -- and the result is that hardly
anyone uses REL. Well-defined semantics for at least some values of
REL would begin to make it useful, and then it just might be taken
seriously by document authors.

Paul Burchard <>
``I'm still learning how to count backwards from infinity...''