Lars (and everybody else listening in but keeping unfortunately silent)

 > >well, current latex already has two for \footnotetext (one being the text
 > >and
 > >one being the number to use or \NoValue)
 >
 > I would rather have thought that the number should be taken from a counter
 > that was managed by the template itself, i.e., the template has one key for
 > name of counter to use and some of the keys which control formatting should
 > (indirectly) refer to this counter. This makes more sense to me, since
 > there is no point in having footnote numbers assigned in any other way than
 > in sequence.

right, that would be the normal case, but you might want to allow to have via
user manipulation offer to manually freeze the counter value. however in most
cases the value passed would indeed be \NoValue indicating to use the internal
counter.

there are some cases where this is useful (and they are not only for technical
reasons)

if one does offer such a possibility (in principle) then it has to be offered
by the template type to be available in each instance.

 > On the other hand, I have now managed to come up with something which, if
 > it is to be handled by footnote templates, definately should become an
 > additional argument to the template. Consider the situation that a LaTeX
 > document containing footnotes is to be typeset as some sort of hypertext.
 > What is then the most convenient (for the reader) thing to do with the
 > footnotes? I would think it is to make links for them (click on the link,
 > get the footnote text on screen). The clickable area of a link in hypertext
 > is, by convention, longer than the one or two characters used for a
 > footnote number, and hence a footnote template that typesets footnotes this
 > way would need an additional "text to make the link text" argument.

that is an interesting one.

let's forget for a second the template side of this, how would you consider
providing an interface for that on the LaTeX level?

this is a serious \footnote[question]{Like this? which would of course be
incompatible with the current interface.}

----------------------------

here are two more which i was wondering if they should get included:

- should refid attributes be passed as arguments, eg instead of the current
 custom in LaTeX which requires something like
   \footnote{\label{foo} ...}
 pass a label  name via one argument to the template (how or if this is
 implemented on the LaTeX user front end would be a different matter)

- more something to solve a technical problem (because it is rather difficult
  in TeX): have one argument that is a boolean and states that the current
  footnote is one in a row of footnotes which belong together
  example:

  text text\footnote{first}\footnote{second}

in most case this would be supposed to come out as

            1, 2
  text text

and not as

            12
  text text

as it would in current LaTeX.

now please don't tell me that \footnote could easily scan for a \footnote
following. i know, but how do i specify this if we separate frontend (ie
\footnote declared via something like the xparse package from the typesetting
declaration)

perhaps the answer is to extend the xparse package to allow specification of
such extended parsing for tokens like \footnote but then again the result of
the parsing would need to be passed on to the processing template and that
would then mean passing something like a boolean to it.

comments? ideas? counter-arguments?

----------------------------

 > [snip]

not getting to the rest tonight (getting rather tired by now)

good night
frank