[Pythonmac-SIG] Which Python are people going to use?

[Bob Ippolito]

On Apr 29, 2005, at 9:03 AM, konrad.hinsen at laposte.net wrote:

> On Apr 29, 2005, at 4:58, Bob Ippolito wrote:
>
>
>> So stop providing libraries for 2.3, eliminate the confusion, and  
>> have them install that one extra package.
>>
>
> That's what I would do if I were reasonably sure that everyone else  
> did the same.
>
> Which makes me wondering: who is distributing Mac Python packages  
> at all at the moment? There's the PackageManager archive, yours,  
> and Fink, but is there anything else? Fink is separate anyway (they  
> have their own Python), and PackageManager seems to be on its way  
> to the museum.
>
> So if your archive is the only one that matters anyway, then we  
> (the participants of this SIG) could start a massive packaging  
> campaign to make it the "industry standard".

I would say that the PackageManager archive doesn't really matter.   
There's not much in there, and from what I've seen it's caused more  
headaches than it has cured.  Also, nearly everything there should  
also be in pythonmac packages.

There's MacEnthon, which matters, and is currently only supporting  
Panther's 2.3.0.  That may or may not change depending on what's best  
for Enthought and their customers/community.  The majority of the  
work that went into MacEnthon was well underway before 2.4.1 was  
available, so it's understandable that they used Python 2.3 at the time.

>
>> Well if it were my decision I'd say that people should stop caring  
>> about 2.3, because 2.4 is better in every way (except "they don't  
>> already have it") and more
>>
>
> Personally, I agree, but then I know how important the "they  
> already have it" (plus "it's the official Apple version") argument  
> is for many others. Since I publish packages for the benefit of  
> others, I do listen to them.

Well, traditionally, the "official Apple version" hasn't been so  
great.  In Mac OS X 10.2 it was a disaster.  Mac OS X 10.3 it was ok,  
but had some problems.  Apple did get everything right in Tiger's  
build of Python, as far as I can tell, however, that only *really*  
matters if you're targeting Mac OS X 10.4 (no more no less) and  
Python 2.3.  I would have to imagine that anything you build that  
depends on Apple's Python 2.3 will no longer function in Mac OS X  
10.5 (because it will hopefully include Python 2.5).  However, I can  
nearly guarantee it WILL work if you stick to a third party Python  
build (unless Apple screws something up badly).

I would recommend use of Apple's Python under the following  
circumstances:

- Writing plugins for otherwise non-Python applications.  There can  
only be one Python interpreter per-process, and it makes sense to use  
the one that's there.  This is a trade-off between cooperation of  
plugins and plugin authors, and compatibility with future versions of  
Mac OS X.  Alternatively, on a per-application basis, it might make  
sense to have a "meta-plugin" that includes a Python 2.4 interpreter  
plus some useful libraries (PyObjC, etc.) that other python-based  
plugins could depend on.

- wxPython 2.5.3 or TclTkAqua applications where using the stock  
libraries and saving 20MB or so (uncompressed) is worth the trade-off  
of targeting Tiger specifically (knowing that Panther and 10.5 will  
be incompatible).  Or otherwise if saving the 1-2MB of interpreter  
overhead is worth the risk, though I'd hope that not to be the case.

- Anything that otherwise targets Tiger specifically where using  
Python 2.4 features isn't important.

-bob