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

Craig Hubley (
Thu, 27 Apr 95 23:04:36 EDT

> > But, the only works if we have an agreed upon semantics. And there are
> > likely other examples, such as TOC, Index, Glossary. A few standard values
> > would go a long way to promoting such use.
> Absolutely. SCO is already using REL with specific keywords --
> as I have mentioned in earlier mail -- and we intend to do more.

So there are link types we might call "SCO/back", "SCO/next", etc. ?
What applications are working now, and are envisioned for these specific

> > There does seem to be a place for standard semantics here. And a place for
> > author defined relationships. Is the REV attribute the right place for
> > both? What about CLASS, which looks to be similar?
> There _is_ a place for standard semantics and standardized keywords.

This has not been proven. More detail on how SCO uses them might help.

> There _may_ also be a place for ad hoc keywords,

My read of the history of hypertext systems since 1967 or so, says ad hoc
author-defined link names seem to be the only universally useful mechanism.

Communities of authoring practice can then evolve, such as SCO appears to
be nurturing, without forcing all the early practices on the rest of us.

My position is the inverse of Murray's: there _is_ a place for ad hoc
link-type keywords with no clear semantics other than what a human can
read into them; Once such a mechanism is in place and clearly working,
there _may_ be a core set of standard types that software can act upon:
a small, well-defined, practice-proven, and precedent-defined subset of
those in use. I would suggest starting with the 'link types' defined by
the NetNews and Mail systems, each of which has a means of unambiguously
specifying other News or Mail files. But avoid words like 'next' that
require a specific ordering on a set of files to be useful...

> but other than
> providing a mechanism for allowing this, there is no place within
> the HTML WG for specifying that.

Of course. However, the 'mechanism' becomes much more complex and
difficult to support in a portable way, as one standardizes deeper
levels of semantics. If we say that a link type makes the browser
behave in a certain way, then we have invented a mechanism to specify
browser behavior, and this will lead us into implementation dependency
for certain.

It seems to me that we have to invent a means of defining types, THEN
we can decide what standard types are required (object oriented dogma).

Anyone who worked on C++ standards can testify that trying to build a
consistent type definition system ON TOP OF a set of pre-existing types
that were standardized before the definition system was invented, is a
horror story that yields results that please no one.

Craig Hubley                Business that runs on knowledge
Craig Hubley & Associates   needs software that runs on the net     416-778-6136    416-778-1965 FAX
Seventy Eaton Avenue, Toronto, Ontario, Canada M4J 2Z5