Agreed. And since very long URIs can occur anywhere, we should
avoid a <FIG>-specific solution. The best practical solution is
probably to make use of element content inside <HEAD>, which is by
default not rendered. In other words, what we need is a non-empty
form of <LINK>, and support for indirect reference to <LINK> URIs.
For the first, perhaps </LINK> could be made omissible, instead of
having <LINK> be empty? That would simplify matters. If not,
invent a new tag like <LLINK>.
For the second, we could either create a new common linking
attribute (extending the current list of REL, REV, etc.), or allow
LINK indirection as I have previously proposed. Link indirection
means: if the target of a link is a another link, follow it too.
So, if </LINK> can be made omissible and link indirection seems
like a good idea, we could use the following mechanism for very long
URIs:
<HEAD> <LINK ID=xxx>http://www/script.class#x=1;y=2</LINK> </HEAD>
<BODY> <FIG SRC="#xxx">...</FIG> </BODY>
The use of element content for URIs does present additional
encoding complications, as you say, but placing URIs outside of the
document would be even worse from a practical point of view.
--------------------------------------------------------------------
Paul Burchard <burchard@math.utah.edu>
``I'm still learning how to count backwards from infinity...''
--------------------------------------------------------------------