15 May 1997 19:50:51 +0100
>>>>> "Henry" =3D=3D Henry G Baker <firstname.lastname@example.org> writes:
Henry> Well, if I am given a random string, I'd be hard put to
Henry> identify it as one which was produced as the result of
Henry> symbol-name. Putting dire warnings into language standards is
Henry> a waste of hot air. If you want a rule enforced, then enforce
Henry> it -- e.g., type the string as 'immutable'.
I completely agree, but isn't that somewhat difficult to achieve in a
language that wants to compete with C in efficiency? I can't believe
that most CL implementors didn't think about the problem. In the
TI-Explorers that kind of error was detected because they had
read-only pages (if I remember correctly). Why most other CL
implementations do not have that?
Henry> "The first thing we do, let's kill all the language-lawyers"
Henry> (apologies to Shakespeare)
Don't take me as language-lawyer. I hope the LispOS will solve many of
the CL problems. But a CL without all the problems you describe will
not be able to run efficiently in most other operating systems.
That's why I think the best idea is not to depend on CL. Our system
language (whatever that may be) should not need to run on top of other
operating systems, otherwise it will have to make commitments
regarding those operating systems.