Re: C as default (was Re: RFC 16 (v1))

[mjd-perl-list-language-strict]

Daniel Chetlin <[EMAIL PROTECTED]>:
> Yep, I agree. I had a short conversation with MJD about this, and he
> made some great points. I don't want to paste it here without his
> permission, but I'm hoping he's listening. Although I do know he's
> going to Las Vegas, so he may not have time to chime in.

OK, here's what I said to Dan:

First, it's silly to talk about 'strict' as if it were one thing.
It's not.  It's three things that have nothing to do with each other.
They should be considered separately.

For example. 'strict refs' forbids the use of strings as references.
Since Larry said on 18 July that probably typeglobs will be going
away, and since that implies a total revamp of Perl's symbol table and
name lookup semantics, it's impossible to say today whether 'strict
refs' is even going to have any meaning in Perl 6.  It's not a
reasonable topic for discussion because the feature it's meant to
disable may very well not exist in Perl 6, or may be altered in an
orthogonal direction so that 'strict refs' is nonsensical.

But that hasn't seemed to stop anyone from shouting 'use strict good!
no strict bad!'  The discussion so far has been disappointing in its
lack of cogency.  The failure to discuss the three 'strict' pragmas
separately is a good example of this.  The title of the RFC itself
perpetuates this thoughtlessness and I think it disables the
discussion.


Second, although I have no opinion at present about the desirability of
'strict refs' or 'strict subs', I think it's a perfectly awful idea to
perpetuate 'strict vars' into Perl 6 at all, and certainly not as the
default.

In Perl 5, variables are global by default and require an explicit
declaration if they are to be lexical.  Everyone recognizes that this
was a design error, so they say 'Always use "strict vars"' which helps
programmers evade the fallout from the design error.  Fine.

Now we are trying to say that it should be the default to avoid the
fallout.  Duh!  Just fix the design error and there will be nothing to
avoid!

The right solution is to make variables lexical by default, and
require an explicit declarations of global variables.  We are not
exactly on shaky ground here.  Here are the names of some languages
you may have heard of that follow this strategy:

        C
        C++
        Common Lisp
        Java

If lexical variables are the default, then 'strict vars' will be
totally unnecessary because it will be impossible to forget the 'my'.
This will be cleaner and safer and more convenient and simpler than
requiring 'use strict' *and* requiring 'my' declarations everywhere
*and* requiring that the programmer get a fatal error if they forget
to put in all 12,393 declarations.