Re: Encoding the name in the file contents
Boris Veytsman <borisv@lk.net> wrote:
> I hate to disappoint you, but this is much more work than you think.
>
> LaTeX is not a Linux project. It is not even a Unix or Posix
> project. It is a thing which works on virtually all platforms
> including Unices, Windows, Mac, OS/2, VM/CVS, VMS, DOS and even Palm
> Pilots. md5sum software is too system-dependent to be included in
> LaTeX.
All of these platforms have C compilers. In fact, I would wager that
all of the platforms that TeX runs on have C compilers. Maybe not
Forth or Pascal or even C++ compilers, but certainly C compilers.
> We cannot presume that a platform LaTeX runs has md5sum
> software or even C available. The only way to use this scheme for
> LaTeX is to rewrite md5sum software or any similar software in
> TeX. This is (a) prohibitively difficult and (b) will slow down things
> on certain platforms up to the point of unusability. Remember, Debian
> does not work on 286 Intels while LaTeX does.
The core md5 program in debian is 241 lines. It includes a header
with all of the various machine dependencies which is 390 lines.
There is also a wrapper script, which the LaTeX project wouldn't need.
Don't tell me that 631 lines of C code is too much. Fiddling with
that program would be much less work than, for example, participating
in this discussion. You probably wouldn't have to modify it anyway.
Henning Makholm <henning@makholm.net> wrote:
> Not really - it would break the impotant technical invariant of
> LaTeX that it is all written in the TeX macro language. I believe that
> this invariant has technical merit: LaTeX should be able to run in
> every platform that has a working Knuth-canonical TeX, even on, say,
> machines with no C compiler.
I assert that there are no such platforms. Please show me
counter-examples.
> A solution that required the LaTeX kernel to be able to interact with
> C-based checksumming code such as md5sum, would itself be a case of
> sacrificing clear technical wins to enforce a social convention -
> exactly the sort of thing that we've used to argue *against* the LaTeX
> people's current renaming principle, just more so.
Actually, the change would be to enforce a technical goal. As many,
many people have demonstrated throughout this discussion, it is pretty
trivial to change things without the user's knowledge.
> Providing an interaction between TeX code and md5sum in the first
> place would not be trivial at all - unless it was done as some sort of
> extremely ugly one-time hack (which would have to be done in the TeX
> engine,
It would be adding a new command to TeX. Something like
\verify_package{foo,(foo's MD5sum)}. Gee, isn't it nice that we can
modify the TeX engine?
> which is presently separate from LaTeX, which is a Good Thing) it
> would open up quite disturbing new security perspectives, for
> instance.
I seriously doubt that there would be any new security problems. The
program is just opening and reading files. TeX already does that.
Yes, any modification to a program can cause security problems, but I
think you're blowing this out of proportion.
However, I'm not going to force this down the LaTeX community's
throat. If they don't want to do it, they don't have to. I just
think that it accomplishes their goals better than anything else,
while preserving the freedom to modify.
Regards,
Walter Landry
wlandry@ucsd.edu
--
To UNSUBSCRIBE, email to debian-legal-request@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
Reply to: