[tex-live] paper size special

[Karl Berry]

    except that all other drivers support it in an incompatible way..

How unsurprising.

    was dvisvgm but I didn't understand its size at all:
    etex pagesize; dvisvgm pagesize
    pre-processing DVI file (format 2)
    processing page 1
      page size: 217.359pt x 664.121pt (76.3929mm x 233.412mm)

I believe dvisvgm trying to create an includable graphic rather than a
full page, thus it's more or less the bounding box of the output?  Full
pages are not what SVG files are for, after all.  The papersize special
is not mentioned on the dvisvgm man page.  Could ask MartinG and/or look
at the source if it matters.

    So... I was wondering if dvi drivers that support special{papersize=
    could agree on how to treat multiple instances?

If the only thing we have to change is dvips, per your experiments, that
is feasible, in principle.  It certainly seems more feasible than
changing "everything else".  And also the behavior (last special on
first page wins) seems better -- matches pdftex better.

    Changing (say) dvips would of course affect the behaviour of any
    document that has this set twice (eg has hyperref and geometry both
    loaded) so it's not necessarily a good idea, 

Let's suppose, hypothetically, that we change dvips to match the others
(with an option so people can get the old behavior if they want it, etc.).

Would this imply you would then make further changes in latex?  If so, what?

And, are any changes in this area desirable at all in latex2e?  I am
skeptical.  Whatever the deficiencies and suboptimalities of the current
situation (no argument that there are plenty), people have necessarily
figured out how to get the page sizes they want with the software they
want.  If we make any changes, I highly suspect existing workflows will
break, and not in any obvious way.

Or am I being too negative?  -k