Message-Id: <9302091259.AA03874@sc.zib-berlin.dbp.de>
Reply-To: Mailing list for the LaTeX3 project <LATEX-L@vm.urz.Uni-Heidelberg.de>
Comments:     Warning -- original Sender: tag was kris@CS.CHALMERS.SE
In-Reply-To:  <9302091158.AA15181@odin.diku.dk>
Date:         Tue, 9 Feb 93 13:38:53 +0100
From: kris@DIKU.DK
Sender: Mailing list for the LaTeX3 project <LATEX-L@vm.urz.Uni-Heidelberg.de>
To: Multiple Recipients of <LATEX-L@vm.urz.Uni-Heidelberg.de>
Subject:      Re: Makeindex reimplementation
Status: R


> I think this discussion of makeindex is rapidly getting into
> dangerous waters.

Yes, definitely.  It should stop now after you all agree with what
I've written below :-)

> I believe very strongly that new components to TeX should be
> implemented in C.  Reasons for doing this.

I oppose strongly...reasons below!

> + C is available on almost every platform in the world. GNU C makes
>   a "free" compiler available to almost everybody. Pascal compilers
>   usually cost a lot of money.

However, TeX is written in WEB.  This means that *really* every
platform with TeX has WEB; this is infinitely better than "almost".
Furthermore WEB is a true high-level language in the sense that it
encourages proper program documentation and a modular programming
style, something that is definitely not true for C.  And WEB *has* an
extensive library in the sense that modules of the existing WEB
programs are easily reused.  WEB is not Pascal, mind you, although the
translation WEB -> Pascal is simpler than WEB -> C.  But both are
available, so you can use GNU CC as the backend (I do).  And you
forget that C is not a true standard!

> + C is easier to port than Pascal. Take a look at the TeX change
>   files for various operating systems. When I last ported TeX to VMS
>   through the Pascal-to-C translator with Unix change files I only
>   had to change the PATH separator character. This means that with C
>   a novice has some chance of getting things right; with Pascal
>   there is no use trying.

This is largely due to the fact that WEB operating system change files
"do it in WEB" whereas the Unix ones just provide a uniform interface
to programs written outside WEB.  This is the choice of the *porter*
as it should be.  Your suggestion will take away that choice.

> + Most implementors use C (I can't prove this). Why make things more
>   difficult than they have to be?

Exactly.  Don't change the source language of TeX.  When I first
ported TeX to a 68000 unix box in '85 I was very greatful that the
support program sources could be preprocessed and prettyprinted using
WEB.

> + There is a good chance that patches can be applied by the users or
>   system managers. Unix utilities like diff, patch and make can
>   automate much of the process. Adding change files to this might
>   break things.

This is solved by a proper WEB programming style -- look at those
change files: they invariable change some section called "System
Dependencies".  The reason "diff" and "patch" are necessary in the
un*x world is exactly the opposite: there is no generally accepted
"proper C programming style".

> I don't think that C is the ideal language. However due to a very
> limited library it is one of the most portable languages available.

I am sorry, but this is wrong.  I have not yet met one single system
that had kjust the standard header files/library functions of the K&R
ANSI C book.

> Please be careful when choosing implementation language. Will the
> selection make it easier for the user or cause her/him to go
> somewhere else for a solution?  Will it mean that I have to endure
> errors that I know how to fix due to lack of the right compiler?

Yes...be careful. Stick with the safe choice: WEB :-)

Regards,
	Kristoffer
______________________________________________________________________
Kristoffer H{\o}gsbro ROSE                     Internet: kris@diku.dk
Chalmers Tekniska H{\"o}gskola, Datalogi       Voice: +46 (031)7725019
412 96 G{\"o}teborg, SWEDEN                    Fax:   +46 (031) 165655