Notes on 2.2's changes

A.M. Kuchling akuchlin at crystal.mems-exchange.org
Thu Apr 11 13:35:41 EDT 2002


In article <Xns91ED5ED8EF7A0lumberjack at 216.148.53.96>, Lumberjack wrote:
> hardship. It is impossible to tell from your list how much _time_ (and 
> money) was consumed in locating and fixing those problems. Basically, you 
> haven't provided data on the more important metric. 

The multi-argument .append() and .connect() calls take a few minutes
to fix once the test suite catches them.  After asking him, Neil said
fixing ptl_compile.py takes around 20 minutes.  Grouch is likely the
messiest case, as Greg may have to spend lots of time crawling up and
down in his code, but I don't know how much time those fixes took.

> And it isn't clear why 
> you felt it was necessary to be compatible with 2.2. Wouldn't it have been 
> simpler to just say "run these on version 2.0.x of Python"? 

This software runs on our Web server, and on servers we control, so we
can run any version we like, if we decide the porting effort is worth
the additional features and bugfixes.  New features sometimes let us
write cleaner code; generators are used in a few places,
unsurprisingly, as are the new email package.  My holy grail is to
someday use a version of the ZODB that doesn't use ExtensionClass,
which seems possible with 2.2.  (ExtensionClass is far, *far* more
annoying than anything that 2.1 or 2.2 has changed; it means we can't
run PyChecker, and can't use __r*__ overloading or and had to work
around isinstance() being broken for ExtensionClass in 2.1.)

The large code base is also a decent test case for Python itself.  We
ran 2.2.1c1 and c2 on our developer machines, though not on our
production servers, and it exposed at least one bug that I could then
fix for 2.2.1final.

> The alleged 
> underlying philosophy seems to be that you can pick whatever version 
> that suits you and stick with it; you are under no compunction to track 
> changes to Python. 
> So why track it if there is no compunction? I'd like to 
> see you address that.

What's to address?  "Because we wanted to" is surely reason enough, if
we're willing to pay the corresponding cost in time.

--amk                                                             (www.amk.ca)
It seemed like the night sucked them up, took them into its dark heart. It
seemed like the darkness swallowed them... Perhaps it did.
    -- The convention breaks up, in SANDMAN #14: "Collectors"



More information about the Python-list mailing list