On Tue, 27 May 2008, Frank Niessink wrote:

> Hi Jerome,
>
> 2008/5/27 Jerome Laheurte <[EMAIL PROTECTED]>:
>> Under Windows and MacOS, nothing. Under Linux, it's difficult to say;
>> considering the various versions of Python/glibc shipped with today's
>> distros, there is no way to make a "universal" binary. The glibc part
>> is moot, since most modern distros use the same major
>> (binary-compatible) version. The Python version may be a problem, but
>> since TC explicitely depends on 2.5, it's not a real problem.
>
> Officially it's python >=2.4 currently, although 2.5 is shipped with
> the Windows and Mac installers.

Indeed. I'll have to check, but I think there is no binary 
incompatibility between the 2.4 and 2.5 C APIs.

>> When I think about it, even for Linux, the prerequisites are not much
>> stricter than what is already required (Python 2.5, latest wxPython
>> version or so).
>>
>> So deploying the binary extension module will work for 99% users,
>> IMHO, without having them install anything else. People for whom it
>> will not work (very old distros users, for instance) probably have the
>> technical knowledge necessary to build the extension from source.
>
> Will building the extension module be a distutils command? E.g. will
> "python setup.py build install" work for Linux users?

Except maybe under MacOS. I'll have to see how distutils handles the 
whole PPC/Intel/universal stuff.

> PS: I'm getting more and more excited about this. We may also want to
> think about making syncing the primary persistence mechanism so we
> could add different back ends that support multiple users more easily
> (e.g. a database).

Mmmh, why not. I still have to study the vcard spec in order to see if 
it allows "private" application-specific data. In the current 
prototype for instance, task parent-child relations are lost to the 
server, among other things.

Reply via email to