Re: patch for 2.1.55 pre-1 minix/sysv/affs

Ralf Baechle (
Mon, 15 Sep 1997 14:23:57 -0700 (PDT)

> While I wholeheartedly agree with all Linus said, I would like to
> make a (minor?) point.
> If I understand it correctly (and I came in contact with Pascal lots
> before TurboPascal was born) the original Wirth Pascal was never
> meant as a "real" or "production" language.
> It was meant as a *teaching* language, aimed to force students to use
> what were *considered* good programming practices.
> If You remember what Djistra said about Basic You can understand the
> point (and line numbering Basic was, with fortran4, the main
> language in non-AI-oriented European universities).
> In this condition a certain amount of "straightjacketing" is
> understandable (and IMHO welcome).
> No one thinks stright[jacketing] Pascal can really be used for
> writing applications (thou certain extensions can; does anyone
> remember Mountain View Pascal? not to speak about Borland).

Older Pascal implementations sometimes other limitations like the
limitation to 64kb pcode/code + 64kb data. Which already was a pain
for serious apps somewhen not too long after UNIX's birth. Which
shows pretty clearly why the language was available on those systems.

> As far as the [in]famous goto statement goes: Wirth knew very well
> it's value. In his Pascal compiler written in Pascal there are a lot
> of "goto 9999" (if memory doesn't fail me) used for error
> escape and recovery.

Which is indeed the only thing for which one can reasonably use goto
in Pascal.

> Dogmatic absolutistic statements are IMHO never worth the paper they
> are written on and, usually, denote a narrow-mindad writer.
> On the other hand, as they say: "along with freedom comes
> responsibility" and responsibility comes with knowledge also.

The thing is that universities and schools today mostly concentrate on
theoretical and practical skills of the subjects they're teaching.
There is a *wide* gap between that and teaching responsibility which
tend more to be part of philosophy.

> I still argue _for_ the *teaching* value of punched cards, which
> *force* students to actually _read_ error messages and _think_ before
> recompiling, but I would never impose such a torture for an extended
> period of time. I find this training invaluable when you came to deal
> with e-mailed bug reports.

Usually (at least and other German universities) this is
today trained with various theoretical machine models like register
machines et al. And I strongly believe that being able to debug without
a debugger is a very important skill. At least software debuggers
a la gdb usually fail when it comes to a certain class of bugs like
cache handling or race conditions in a kernel.

Btw, I use the backside of old, leftover punchcards as notepad. No holes
in 'em, that is :-)