A counter-counter-proposal for PEP 236: #pragma( nested_scopes)

Tim Peters tim.one at home.com
Fri Mar 2 18:41:42 CET 2001


[Tim]
> Hmm.  Maybe that's why I so dislike Martin's attempt to hijack
> PEP 236:  his current formulation of an alternative leaves only the raw
> names, without the __future__ or __possible_world__ qualifiers.

[Martin von Loewis]
> If *that* is the complaint, then the patch can be easily changed
> into reading
>
> directive future nested_scopes
>
> (future would be the directive name, and nested_scopes the atom
> argument).

That would be much better to me.  You need to write a PEP for this, though!
As is, "the patch" is a moving target and it's not at all clear how you
intend for this gimmick to evolve.  For example, in other msgs I read today,
you're saying that "directive" *will* be a keyword, someday -- I had not
guessed that (to the contrary, the emphasis before today was on that
"directive" was *not* being made a reserved word).  It needs a spec.

> In that case, I'd prefer /F's interpretation of the state
> of things, i.e.
>
> directive experimental nested_scopes
>
> since this does not talk about the future, but about the present (just
> as the future import does not talk about the future, but about
> directives that are presently needed).

nested_scopes talks about the future because, barring an unexpected disaster,
nested_scopes WILL BE the rule in 2.2, but ARE NOT in 2.1. "will be but are
not" == "the future", wrt any release (incl. 2.1) in which nested_scopes are
not yet the rule.  "will be", "not yet", ... it's impossible to tell the
nested_scopes story without invoking the future!  "experimental" misses the
point:  this isn't an option, this isn't going away, it's not up for a vote,
it's a preview of Python's future.

"experimental" would make more sense for a "hmm, Guido can't make up his mind
whether Python wants this or not, so we're enabling it as an experiment --
let us know what you think!" kind of gimmick.  The only good example of this
in Python's history I can recall is the now-defunct "access" statement (which
few people knew existed, because Guido was so wary of it that it was never
documented -- but that meant it didn't get a good trial, either).

"experimental" is a "use at your own risk" thing; "future" is a "*don't* use
at your own risk, because in a future release you'll be forced to live with
it" thing.  Character encodings are neither of those, they're a "use or not,
it's up to you, the language will support whatever decision you make, today
and forever more" thing.  Those distinctions are crucial for users:  they're
the differences among facing the reality of personal mortality, trying Perl
for a week, and deciding whether to hit the snooze button when the alarm goes
off <wink>.

without-a-pep-it-ain't-hep-ly y'rs  - tim





More information about the Python-list mailing list