why is e.g. web2c in the texmf tree?

[Ulrik Vieth]

> I'm quite sure that this has been discussed extensively, but I could not
> find a hint for the very reason.

I think it's explained in one of the appendicies "Unspecified pieces"
where it says that binaries or implementation-specific files might 
be located either outside the TEXMF tree or in a platform-specific
subdirectory within TEXMF.  teTeX-0.4 uses TEXMF/web2c, and the same
applies for web2c-7.0 as described in another appendix "Examples of
implementation-specific trees".  If E. Mattes suddenly came along 
with an official recommendation for the organization of TEXMF/emtex, 
we'd be happy to add another subsection, but if he prefers to keep
his binaries/formats/bases outside the TEXMF tree that's OK as well.

> My goal is to build a common texmf tree for an emTeX and teTeX 0.4
> installation, and the web2c directory contains material which is
> definitely teTeX specific (I use a second texmf tree to overcome
> this problem) having extensions longer than 3 characters etc.

I don't think it's necessary to worry about the web2c subdirectory
if emTeX never looks into it.  Likewise, you could add a texmf/emtex
subdirectory alongside and teTeX wouldn't bother about it either.
I really don't see the problem.

> Another question which is also related to the TDS: will web2c-7.0 support
> DOS filename searching? E.g. LongTable.sty: first you search
> LongTable.sty, then longtable.sty, then longtabl.sty (8.3), and finally
> longtble.sty (5+3.3). Then you could access TDS trees on CDROMs without
> great problems.

Well, personally I'd prefer to go with RockRidge CDs on Unix systems
and leave it to the DOS programs to get along with 8+3 truncated
names, if they just see plain ISO-9660 file names.  As long as the
truncated names are unambigous within a directory, "longtable.sty"
should indeed be stored as "LONGTABL.STY" on ISO-9660 which should 
be found using the search strategies currently implemented in emTeX 
when \usepackage{longtable} is requested.  Of course, the question 
arises what'll happen, i.e. how the file names would show up, if 
a Unix filesystem containing long names is mounted on a DOS system. 
Does it just truncate long file names or does it mangle them?

Anyway, I wish you success with your project to integrate emTeX 
into the TDS.  If you succeed, it might be interesting to consider
integrating it into the next edition of the TeX Live CD which is
currently probably of little use on DOS systems, although it should 
ultimately become a multi-platform TDS distribution not limited
to Unix systems.
 
Cheers, Ulrik.