>> #{h1=1},{h1=2}
>
>But people don't (generally) want to jump to arbitrary points in text --
>they want to jump to points which have *meaning* and that meaning can often
>be expressed within an encompassing scope.
Very true.
>Positional markers are completely luzing in that whenever you change the
>text, your markers go to heck. It would be nice to have markers that could
>survive across versions of docs if their (intended) targets survived.
Again, very true.
>It seems like a useable way to do sub-document reference would be to
>implement some form of tag on HTML markers so that they could be pointed at.
>Servers could then send back parts of docs, rather than just blasting whole
>schmears down the pike.
Well, adding an attribute, or perhaps using an element as a
marker, could be very nice, except that you have to maintain
the identifiers in some way. Positional markers lose only if
you assume that the URL's using them are static. In most
situations where sub-document naming is desired, it is quite
feasible to generate the relative sub-document addresses on
the fly. For example, when a user opens a "book", a TOC could
be sent over. The TOC would contain section titles with URL's
using a relative address, generated on the fly, in the hyperlink
to a lower level. One advantage to such a scheme is that it allows
quick access to the particular element in the SGML tree that
you want.
Of course, another way would be if would could simply type in
some thing like:
"book about fir trees chapter dealing with pollen"
rather than a URL....