[tex-live] Kpathsea in TL2007

[Heiko Oberdiek]

On Sat, Feb 17, 2007 at 05:19:53PM +0100, Zdenek Wagner wrote:

> 2007/2/17, Heiko Oberdiek <oberdiek at uni-freiburg.de>:
> >On Sat, Feb 17, 2007 at 10:09:57AM +0000, Philip & Le Khanh wrote:
> >
> >> Akira Kakuto wrote:
> >> >
> >> >I forgot to write about a small difference between
> >> >Windows and other systems concerning the order of
> >> >file search.
> >> >
> >> >[1] Windows
> >> >    (a) \input foo
> >> >        try foo.tex first, if not found then try foo
> >> >
> >> >[2] Other systems
> >>
> >> >    (c) \input foo.bar
> >> >        try foo.bar.tex first, if not found then try foo.bar
> >>
> >> The behaviour as described for Windows seems logical (indeed,
> >> required) to me : imagine one has
> >>
> >>       \input letter-spacing.sty
> >>
> >> then surely one would want letter-spacing.sty to be
> >> loaded, and not letter-spacing.sty.tex, even if the
> >> latter existed.
> >
> >But (a) prevents that "foo" can be loaded if
> >both "foo" and "foo.tex" exist. :-(
> >
> >Therefore I prefer a new \input primitive that uses
> >curly braces (such as \pdfximage) without default
> >extension. Then it would be possible to reliable
> >address each file name (including spaces, quotes, ...)
> >
> It is not a good idea because \input{foo} is a LaTeX macro and assumes
> to load foo.tex if it exists. It can make things confusing.

No, it isn't, I said a *new* primitive, of course with a new name, e.g.
\inputfile, \pdfinput, ... or whatever name.
Also the existence of this new primitive can be tested in a reliable way.
Thus the compatibiliy isn't violated and new macro code can use
the new primitives. It reduces confusion because nobody must
guess which file names can be detected and which files are loaded.
(Also new primitives for \openin/\openout are needed.)

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