How important is Python 1.5 compatibility?

[Erik Max Francis]

How important do people think Python 1.5 compatibility is? For some of
my projects I've strived to keep it compatible with 1.5 (at least for
the portions of the applications that actually would work with 1.5; for
instance Unicode support obviously requires Python 2.x). It seemed to
me that this used to be fairly important, when, say, 2.1 was new, mostly
because of other large Python applications that required 1.5, or lazy
ISPs who still used older version.

Based on the discussions I've seen, I'm getting the impression that this
is less and less the case as time goes on (which is, of course, exactly
how you'd expect things to be). How important is Python 1.5
compatibility today?

 

[Christopher T King]

me that this used to be fairly important, when, say, 2.1 was new, mostly
because of other large Python applications that required 1.5, or lazy
ISPs who still used older version.

I know of at least one ISP that still uses 1.5.2.

Based on the discussions I've seen, I'm getting the impression that this
is less and less the case as time goes on (which is, of course, exactly
how you'd expect things to be). How important is Python 1.5
compatibility today?

If you're writing web stuff, I'd try to be 1.5 compatible for those lazy
ISPs. As to desktop programs, however, I usually don't even attempt 2.2
compatability, since Linux installations are usually already at the latest
version, or easily upgradable, and on Windows, you can distribute your app
with whichever version you like using py2exe. Correct me if I'm wrong, but
I believe the latest version of OS X comes with 2.3, too.

 

[Roger Binns]

How important do people think Python 1.5 compatibility is?

Heck, I am just dropping support for Python 2.2! I don't think
there are any supported Linuxen (other than RHEL which I don't care
about) that use anything less than Python 2.3. You can trivially
install the pydotorg 2.3 alongside any existing system Python
anyway.

Roger

 

[David Bolen]

Based on the discussions I've seen, I'm getting the impression that this
is less and less the case as time goes on (which is, of course, exactly
how you'd expect things to be). How important is Python 1.5
compatibility today?

I probably have to answer that in two ways for myself:

First, while our new development projects are holding Python 2.2 as
the baseline (probably move that to 2.3 before they get released),
Python code in our production deployed systems (a few thousand
machines at ~500 sites) is still entirely on 1.5.2, as is our
production central services for those sites - although the latter not
changing is more a lack of need than the logistical hurdle that the
sites represent. So I certainly still appreciate and prefer modules
that are 1.5 compatible when possible.

However, even with such an installed base, I don't have my head in the
sand and don't necessarily expect most current packages to continue to
support 1.5. At the moment, in most cases when a package that I need
doesn't, I can continue to run those sites with an older version of it
that did at one time support 1.5. That's probably part of the reason
I haven't taken on the operational overhead of upgrading yet.

So I can't say that I'm against packages enforcing a more modern
minimum (that's probably 2.2 in my mind at the moment). My guess is
somewhere in the next year or two I'll really want some particular
package and take on the upgrade task to transition my installed base
to a 2.x series.

-- David

 

[Peter Hansen]

How important do people think Python 1.5 compatibility is?

IMO, absolutely unimportant. I'd ignore it completely.

I keep 2.2 compatibility in mind at this point, but nothing earlier.

-Peter

 

[Josef Dalcolmo]

Heck, I am just dropping support for Python 2.2! I don't think
there are any supported Linuxen (other than RHEL which I don't care
about) that use anything less than Python 2.3.

well, Debian stable is still Woody and still uses Python 2.1.

- Josef

 

[Roger Binns]

well, Debian stable is still Woody and still uses Python 2.1.

I guess that would be the other exception. However I was
under the impression that Woody users were in two classes:

- stable systems that are not touched and do not have
new software installed on them

- systems where new software is installed and many packages
from testing are used

The former category is less relevant since they are unlikely
to install new code you are writing, and the latter category
can run Python 2.3 alongside 2.1, or replace it completely.

So the versions pretty much look like this for users who
have not upgraded their systems nor updated their Python
version.

Python 1.5 Redhat 7.3 and before (no longer supported)
Python 2.1 Debian Woody
Python 2.2 Redhat 8 and 9 (no longer supported)

Roger

 

[Christopher T King]

well, Debian stable is still Woody and still uses Python 2.1.

But who ever uses stable?

 

[Josef Meile]

How important do people think Python 1.5 compatibility is?
I think one as developer has to understand that the languages are
evolving every day, so, one has to update the code as well and take
advantages of the introduced features.

For example, consider working with strings:

in python 2.x you can say:
pos=myString.find('test')

on 1.5.x you can't do that. You have to do:

from string import find
pos=find(myString,'test')

I find the latest sintax is ugly and not object oriented. For style
matters, I would switch to python 2.1.x

Regards,
Josef

 

[John J. Lee]

How important do people think Python 1.5 compatibility is? For some of
my projects I've strived to keep it compatible with 1.5 (at least for
the portions of the applications that actually would work with 1.5; for
instance Unicode support obviously requires Python 2.x). It seemed to
me that this used to be fairly important, when, say, 2.1 was new, mostly
because of other large Python applications that required 1.5, or lazy
ISPs who still used older version.

[...]

Last time I checked (a month or so ago), the sourceforge web servers
were still running 1.5.2. The SF *shell* servers were more
up-to-date, but not the web servers.

I take an "I've started so I'll finish" attitude to
1.5.2-compatibility with my open-source stuff. With code where I
started out supporting 1.5.2 (um... badly, at least before I had
decent tests), I feel I may as well leave it that way. (despite the
fact that I'd be surprised to learn that anybody actually runs my code
on 1.5.2)

With new open-source library code, in the absence of a need for some
major feature introduced later, I'd probably pick 2.2 now, because of
the big changes from 2.1 to 2.2. I can see why you might choose 2.0,
2.1 or 2.3 instead, though. All depends on the audience you hope to
reach, of course.


John

 

[Richie Hindle]

[Erik]

How important do people think Python 1.5 compatibility is?

We still have 1.5.2-based systems in production use, for various reasons.
One is "if it ain't broke, don't fix it." Another is that some are on
obscure platforms and getting Python to build there was some effort, which
we don't want to gratuitously spend again. Another (as Christopher pointed
out) is that some ISPs are still running 1.5.2.

 

[Aahz]

How important do people think Python 1.5 compatibility is? For some of
my projects I've strived to keep it compatible with 1.5 (at least for
the portions of the applications that actually would work with 1.5; for
instance Unicode support obviously requires Python 2.x). It seemed to
me that this used to be fairly important, when, say, 2.1 was new, mostly
because of other large Python applications that required 1.5, or lazy
ISPs who still used older version.

Based on the discussions I've seen, I'm getting the impression that this
is less and less the case as time goes on (which is, of course, exactly
how you'd expect things to be). How important is Python 1.5
compatibility today?

As of yesterday, we (at my new job) just dropped the need for 1.5.2
because we dumped the old server. It's not clear whether our target
will be 2.2, 2.3, or 2.4 for the future; I'm going to argue for 2.3
if there's any intent to use new-style classes. (The new server is
running RH9 with Python 2.2 as the default.)

 

[Jorgen Grahn]

well, Debian stable is still Woody and still uses Python 2.1.

Actually, Python 2.2 is also in Debian Woody (although 2.1 is what you get
by default if you just generally tell it to install "python").

/Jorgen

 

[Michael Hudson]

With new open-source library code, in the absence of a need for some
major feature introduced later, I'd probably pick 2.2 now, because of
the big changes from 2.1 to 2.2. I can see why you might choose 2.0,
2.1 or 2.3 instead, though. All depends on the audience you hope to
reach, of course.

I can't think of a good reason for choosing 2.0. Or, if you're
choosing right now, 2.2, really.

Cheers,
mwh

 

[Michael Hudson]

[Erik got a personal copy by mistake]

How important do people think Python 1.5 compatibility is?

At the ACCU conference a few months back, I was talking about
new-style classes and asked the audience what version of Python they
were using. I was surprised by the response: apart from one guy who
said '2.1 when I'm forced to work with old versions of Zope'
*everyone* was using 2.3.

Cheers,
mwh