[tex-live] Bad .cfg files in TeX source tree?

[Heiko Oberdiek]

On Wed, Feb 07, 2007 at 09:49:28PM +0100, Frank Küster wrote:

> Jonathan Kew <jonathan_kew at sil.org> wrote:
> 
> > Packages like this that test the engine and configure their back-end
> > accordingly need to be updated to support xetex; in many cases, this
> > will just mean checking for the xetex engine (e.g. using ifxetex.sty)
> > and activating the same code as for pdftex. I haven't had time to
> > pursue this with all the package maintainers, though (and until xetex
> > became part of mainstream distributions, the incentive to support it
> > in standard packages wasn't so strong, either).
> 
> Wouldn't it be a good idea to have this functionality at a central
> place, i.e. one file that's loaded by many packages?  This would also
> give users a central place for switching the default for DVI output from
> dvips to dvipdfm(x) or hypertex.  Currently there are a couple of
> packages which hardcode dvips (Today I came across one that loads
> [dvips]{hyperref} if \pdfoutput=0).
> 
> Then we would have to convince package authors to use this interface to
> the detection of engine/output format, and both distributors, local
> admins or individual users could make sensible choices.

Unhappily it is not trivial:
* For example, you setup a central place with "hypertex" as
  driver for DVI output. Package hyperref will be happy, package
  color or graphics won't.
* Granularity:
  * Most packages don't distinguish between dvipdfm and dvipdfmx.
    Unhappily there is a small difference that is hit by package
    bmpsize. Thus the package must differentiate.
  * Or consider package dropping. The right driver option for
    pdfTeX in PDF mode is dvips(!).

Yours sincerely
  Heiko <oberdiek at uni-freiburg.de>