Warning switches (was: Python 1.6 alpha 1 released)

[Johannes Stezenbach]

Fredrik Lundh <effbot at telia.com> wrote:
>Johannes Stezenbach <yawyi at gmx.de> wrote:
>> Will python then support a -w switch to support people in writing
>> correct programs (i.e. compatible with likely furure language/library
>> changes), just like our-favourite-four-letter-P-language?
>
>how about -v<version> ?
>
>there has been talks about compatibility "modes" (probably an option,
>but probably also a way to set options programmatically), as a way to
>help people prepare for "Python 3000".  I think we might expect some
>of it in 1.7, and more of it in 1.8 (if that ever materializes).
>
>but hopefully, Python 3000 will be completely free from modes.

The next word I had written should have alarmed you that I was joking:

>> Seriously:

I really don't want -w. -t is already enough for me.


>> Just postponing the change is evil since it will only hurt more people
>> in the future, as the number of python users is growing. (CP4E?)
>
>yeah, but at the moment, guido and the rest of the 1.6 team have
>more important things to do than to deal with the spanish inquisition.
>so I'm afraid it'll have to wait until that -v1.6 switch is in place.

Well, this thread certanly got a bit lenghty but I wouldn't call
it "the spanish inquisition". And even if Guido's mailbox overflowed
from complaints, what's there to gain in postponing the connect()
change? He will just get all the complaints again in 1.7, and in
numbers which increase with the number of python users.
And I don't think that claiming "connect(host, post) is deprecated
and won't work in 1.7" in the documentation is a Good Thing,
because there is a lot of code "out there" (e.g. python.org,
parnassus, the starship, ...) which people will download and
use as examples or building blocks for their own stuff -- and
it's all broken and will stop working in 1.7!
My point is simply: Do it now or never [1]. The longer you wait the
worse it gets.

And just to waste some more bandwith let me repeat (yawn) what has
been said before by others in this group:
a) There are much more changes in python 1.6 which will break user code.
b) The socket.connect() stuff does most likely affect only a few
   lines for most python users, and these are most likely easy to
   find and fix.
c) It's still months until 1.6final will be out.
d) No one is forced to upgrade immediately.

>> If it's not broken then don't fix it.
>
>    "I try to avoid breaking user programs, but if something
>    needs to be done it really needs to be done."
>    -- linus torvalds

Again I feel you misunderstood me, maybe because I snipped too much
of Guidos posting. I object to Guido "fixing" the code so that
the broken 2-arg connect() syntax becomes valid again. I'm more
with Linus' quote.

-eot


1) The "never" option would be OK for me. I'm not a language lawyer.