The original issue of having a way to define a hyperlink/button in the body
of the document, that acts like a browsers "Back" button does not seem to
fit the model of the many other previously suggested REV/REL
realationships. Most of them seem to concern static relationships between
documents/URLs, this is an operation that makes assumptions about the
browser having a history list/stack.
The suggestion about defining a URL scheme for local browser behavior has
the feature of dumping the issue in a different document, and maybe a
different working group, and removing all SGML issues. It would be easy for
implementions that do not understand it say so.
But, on the other hand, it is somewhat worse that "news:" in being a very
non-universal scheme, and seems prone to ad-hoc or dangerous extensions.
(i.e. "Blink Logo","Send Apple Event","Pay $5 NETCASH")
I'd imagine one way to implement this kind of thing on the SGML side is to
allow markup of the form:
<a NEW-ATTRIBUTE=NEW-VALUE>text text</a>
to create a hyperlink that does something context-specific with the
browser. (Note the absence of other attributes)
We could also say that:
<a rel=VALUE>text text</a>
would define a hyper-link/button in the absence of an HREF for _some_
relationships. The existing writeup seems to suggest that it and <LINK>
define characteristics of jumps defined in terms of HREFs.
Both these might do odd things to existing browsers.
I don't want to ask/require that a browser overrride its own local/history
functions with absolute jumps or other odd notions of the document author.
Having ways to navigate/use relationships defined in the document, or
invoke local functions from a document seems plausible but hard to agree
on.
The stuff necesary to print a tree of documents according to a specified
threading could be a non-trivial real-world application of a good
definition of link semantics.
---
Albert Lunde Albert-Lunde@nwu.edu