🚀 go-pugleaf

On 11 Sep 92 07:46:44 GMT, rolfl@karl.uio.no (Rolf Lindgren) said:
Nntp-Posting-Host: karl.uio.no

rolfl> (Ajay Shah) writes:

Shah> I find it hard to appreciate why TeX should not know how to
Shah> dynamically allocate RAM on demand.  Does this have something to
Shah> do with being written in pascal?  I seem to recall that new()
Shah> returns nil if you are out of memory..

Shah> In any case, where is TeX implementation going in the future?

rolfl> When Knuth wrote TeX, he knew what he was doing. This is a
rolfl> fundamental assumption, and all arguments pro or con should be
rolfl> checked against this.

rolfl> Now, Web, I suppose, supports dynamic allocation of memory
rolfl> through Pascal's new() operator, whose behavior in extreme cases
rolfl> is unspecified and the execution speed of which is compiler
rolfl> dependent.

Web is not the problem. Knuth wanted TeX to be maximally portable, so he
wrote it in Web and in a *subset* of Pascal, one that he thought would
be more portable than full Pascal.

For some bizarre reason he thought that new() would not be available
everywhere, so he avoided its use, and the TeX implementation is based
on the assumptions that each record is really an entry in some fixed
size table. This is terrible, not just in terms of flexibility (none...)
but also in terms of paging, because of the separate, fixed size, memory
pools for each record type (on the other hand it helps when one wants to
\dump, as one just writes out the contents of those tables, without
having to traverse pointer chains).

There is a person, Pat Monardo, who has been working on a
reimplementation in C of the entire TeX program, one that does use
malloc() for allocating record types. I have worked a bit on it,
towards the end of making its allocation of memory even more dynamic,
but I had to stop that for the past year or so. In the meantime
the author has published release 3.3 of his CommonTeX, which is
a high quality program, available as:

	isis.cshl.org:/tex/commontex330.tar.Z

I really hope at times that Knuth's TeX just went away. Pascal is
passe`, the subset he used is too restricted, and literate programming
may not be that great an idea...
--
Piercarlo Grandi                   | JNET: pcg@uk.ac.aber
Dept of CS, University of Wales    | UUCP: ...!aber-cs!pcg
Penglais, Aberystwyth SY23 3BZ, UK | INET: pcg@aber.ac.uk