[tex-live] kpsewhich

[Hans Hagen]

At 13:33 27/03/2004, George N. White III wrote:
>On Fri, 26 Mar 2004, Olaf Weber wrote:
>
> > What you need for this is to use an interpreter that can, in a way
> > that is common to all its incarnations on different OSes, be
> > instructed to start _this_ command with _these_ arguments, and have
> > the arguments treated as literal strings.  If the interpreter doesn't
> > allow for that, than _it_ cannot be used for OS-independent work that
> > involves calling helper programs.
>
>Interpreting command-line args is not the only difference between
>the various shells (*n*x *sh, command.com, cmd.exe, ???).
>If you are going to start defining the requirements for a suitable
>interpreter, don't forget to include calling a helper program
>with an explicit path -- texexec.pl was broken on Win9x because
>it tried to use "/" in explicit paths.
>
>TexLive currently follows a unix-style approach that relies on many simple
>programs working together.  Windows ports of these utilities don't always
>deal with the differences in a consistent way, you can't be sure what
>programs are installed on a given machine, and there are performance
>issues.  fpTeX already provides perl and other tools that come with *n*x.
>Maybe TL needs to provide a shell (one of the reasons our users loved
>4allTeX).

i think that today's users expect integration and that is indeed 
accomplished by 4tex like environments; this means that the tools that make 
up tex should be as platform indepedent as possible and provide 
alternatives for bypassing shell escaping and expansion; many simple 
programs working together is not a problem, as long as they can communicate 
without interference from shells; if shell-like behavior is expected, then 
it should be built into the utilities

Hans