Re: [9fans] samuel

["Thomas Bushnell, BSG" <tb+usenet@becket.net>]

geoff@collyer.net (Geoff Collyer) writes:

> Assigned gotos, COBOL "alter" verbs, writing large programs in
> assembler, self-modifying code, and many other things in this young
> field are bad ideas, but they aren't archaic, even if they are
> wretched mistakes.  Old != bad, and New != good; quality and age are
> largely unrelated, though with luck we learn things with experience.

I think what makes them archaic is that they are now moribund: people
recognize they are mistakes, and they date from the very early days of
the field.  Being bad isn't something automatic with archaic.  But
something still in use isn't archaic; the reason those things aren't
still in use (or shouldn't be ::grin::) is that they are now seen bad.

So not "bad because archaic".  And not really "archaic because bad".
Rather, "archaic" meaning "dating from the earliest days, and not
current now", and it happens that those techniques are no longer
current because they happen to be bad.  A little reflection, however,
will show that being bad is neither a sufficient nor a necessary
condition for a programming technique falling into disuse.  So there
are many archaic things which are not bad (like, say, nine track
magtape) and there are many bad things which are not archaic (like,
say, FORTRAN, which is old but still in use, and perl, which is new).

"Archaic" is relative to the context.  In my usual circles, "archaic"
means "before ancient", where "ancient" refers to something roughly
like 500 BC - 400 AD.  But surely in CS, archaic means "dating back to
the earliest days of computing", and ancient "slightly newer than
archaic".

Borrowing the usual sense from classical studies, might mean that we
should call Babbage and early APL and lambda calculus "archaic", and
we should call things like computed gotos and self-modifying code
"ancient". 

Thomas