[Zope3-dev] Re: Requiring Python 2.3

Martijn Faassen faassen at vet.uu.nl
Thu Sep 25 07:13:42 EDT 2003


Tim Peters wrote:
> [Martijn Faassen]
> > Tim Peters wrote:
> > [using 2.3 is the sane, obvious thing to do]
> 
> For a developer judging by technical merit, or for someone who values
> cost-free support done by others, that's so. 

Agreed.

> There may well be other
> considerations, but since I'm a a developer judging by technical merit, I'm
> incapable of considering them <wink>.

Naturally. :)

> > In this case it is important that Zope 2 gets a 2.3 stamp of approval
> > as soon as possible.
> 
> No released final version of Zope 2 even has a 2.2.3 stamp of approval yet,
> right?

Exactly. It's taken too long, which is why I would urge a 2.3 stamp of approval
landing on Zope 2 rather more quickly than the process for 2.2 has taken..
Especially if Zope 3 will depend on 2.3.

>  Since the 19 months of Python 2.3 development after 2.2 was released
> went by without stamping 2.2, and now a few more months have gone by and
> 2.3.1 is out (2.3.1 was released Wednesday), it makes a whole lot of sense
> on many counts to skip the Python 2.2 line.

On the other hand this amount of time has allowed 2.2 to filter through to
all kinds of Linux distributions and a lot of modules to be available for
it. 2.3 hasn't had that time yet. Then again, it wouldn't matter that much
for a typical Zope 2 setup, as either a binary python is included or the
situation is so complex that maintaining a self-compiled python 2.3 version
is of a minimal concern.

We'd need to ask on Zope 2 related lists to be sure this is true, though.

> > I'm not running Zope 2 on 2.2 yet even as Zope 2.7 isn't out yet
> > which has official support for 2.2 whereas Zope 2.6 still blathers if
> > you start it with 2.2.
> 
> Many Windows Zope users must have already bitten this bullet, since no
> version of Python 2.1 supports "large files" on Windows (full support for
> that didn't exist before Python 2.2.3, where it was backported from
> long-working code new in 2.3).

Windows Zope users typically use the binary that comes with Zope itself and
don't replace it, I'd expect. The typical Windows Zope user likely doesn't 
need such large file support anyway. Anyway, I'm a bit loath to tell a
customer "Oh, you can run it with 2.2 just fine, it just complains at you
for fun". If something goes wrong you can bet the Python version will
become under scrutiny, even though it may have nothing to do with it.

Actually this would be an argument for removing the complaining message
altogether.

> > Let's try to avoid such a situation for 2.3.. Since 2.3 is not a very
> > end-user visible change (unlike 2.2) this should be relatively easy
> > to accomplish if 2.2 support already exists.
> 
> That's what Shane and Tres have testified about their Zope-under-2.3
> experiences, and, indeed, the Zope2 and ZODB3 and Zope3 and ZODB4 test
> suites were ones we routinely ran under CVS Python during 2.3 development.
> If there's something that still doesn't work under 2.3, it must be something
> not covered by the test suites.
> 
> I've got another reason for wanting to skip 2.2:  there's an excruciating
> subtlely involving the interaction between new-style classes and ZODB's C
> persistence code that leaves zodb33-devel-branch working fine under 2.3 but
> segfaulting left and right under 2.2.3.  We have more work we want to do on
> ZODB and ZEO than we can do, and I would rather spend 3 days improving ZEO
> scalability than hacking in layers of painful tricks to worm around the
> relevant flaw in Python 2.2.3.  So that's something to keep in mind too,
> easy to overlook if you're not doing the work:  it's not a question of
> whether you get 2.2 compatability in addition to everything else, it's that
> you get 2.2 compatibility at the expense of not getting other things.

Agreed. I tell my customers the same thing when they want us to develop
features for older versions of our product.

Anyway, moving along to 2.3 and dumping 2.2 sounds like a reasonable plan
to me. My main concern is that I don't want to wait another year before
Zope 2 is declared ready for Python 2.3, as I hope to start deploying Zope 3
technologies within Zope 2 sometime next year.

Regards,

Martijn




More information about the Zope3-dev mailing list