staschuk at telusplanet.net
Wed Jun 11 04:20:32 CEST 2003
Quoth Fredrik Lundh:
> (5) Python programmers don't have anything important to do with
> their lifes, so they're happy just changing perfectly working code
> over and over and over again for every new release, just to satisfy
> the whims of some random comp.lang.python poster.
With sincere respect, the sarcasm is a bit excessive, and the
implications insulting. I do not formally propose backwards
incompatibilities on a whim; neither am I indifferent to the needs
of the existing user base.
Now, to the point. Consider this schedule: this year we revise
PEP 8 to recommend explicit instantiation; in 2005, we formally
deprecate implicit instantiation in the documentation; in 2007 we
add warnings; in 2009 they become errors. Unacceptable? How
about if those two-year gaps are replaced with four-year gaps?
At some point the expectation must be that virtually all code has
been rewritten anyway, and so a minor change to exception raising
syntax does not increase costs at all. Somewhere between that
point and the schedule presently proposed in the PEP (which I view
as a lower bound, the minimum conceivable timescale for the change
proposed), there lies a point at which the migration pain equals
the benefit (unless one thinks the benefit is nonpositive). Any
schedule longer than that should be fine.
So - 3.0 too early? Ok. Tell us what you think a reasonable
schedule might look like. Constructive feedback, that would be.
Steven Taschuk o- @
staschuk at telusplanet.net 7O )
More information about the Python-list