[texhax] (no subject)

[Robin Fairbairns]

phil taylor writes, quoting Herbert Gintis:

> >         I totally agree that it is an unbelievable coup that
> > Knuth's software of the 1970's survives to this day and remains
> > quite excellent. However, the language he wrote it in is a
> > complete mess. All the \expandafters and \csnames and Lord knows
> > what crazy commands, the absence of real floating point, horrible
> > debugging instruments, not object-oriented, etc. I really hate to
> > hack this code, and would love it if TeX were completely rewritten
> > in C++, Delphi, or Java conforming to modern programming
> > conventions.
> 
> There's a strange convolution/conflation of ideas here; the 
> language in which Don /wrote/ TeX (82) was Web, based on
> standard Pascal.

indeed, the idea that one would gain anything going from pascal to c++
seems to me peculiar in the extreme.

> It has none of the "crazy features" to
> which you refer.  However, the language that Don /created/
> (i.e., the TeX (82) language) is indeed quirky, and there 
> there are rumours that one reason for this was that these
> quirky ideas had previously failed to gain acceptance amongs
> his peers, and he saw TeX as a way of demonstrating their
> validity.  It would be interesting to know if any who were
> around at the time (Pierre, Nelson, Barbara, ...) have any
> recollection of the truth or otherwise of this rumour.

i doubt it.  we had a general-purpose macro processor in the late
60s/early 70s, called ml/1, many of whose oddities came back to me
when i started learning tex.  i suspect a macro language is pretty
much inevitable if you have ambitions like knuth's, so the nature of
the language may itself be inevitable.

> >         By the way, I have always found it curious that Knuth is
> > such a great computer theorist and a quite good mathematician, but
> > his computer code (e.g., from his multi-volume book on computer
> > algorithms, or his web, tangle, weave, etc.) is really quite
> > awkward and highly inelegant. A good computer program should look
> > like a haiku poem, not a tangle of impenetrable code, IMHO.

yeah.  in os research, i used to get all worried about the size of
critical routines: i used to turn out real beauties, but i was always
_sooo_ slow.

> Hmmm, just seventeen syllables, with a fixed metre; 
> probably best done in Perl, then !

great.  impenetrability combined with economy: just like my critical
routines.