The BOINC integrated Python interpreter is working and is composed of the
following:

- An executable (pyboinc[.exe])
- Associated libraries (.dll/.so)
- The standard Python library (.zip)

The only question now is, how do we send the standard Python library and
include optional 3rd party libs? Here are two possible ways:

Option 1) Package the base library (python25.zip) with the science app;
workunits must specify additional 3rd party libs (i.e. NumPy10.zip) that are
needed for execution. The specified dependencies would be loaded by PyBOINC
when the job is executed. All 3rd party libraries would need to be installed
as part of the science app to support platform-specific dependencies (for
example NumPy uses C libraries for speed).

Option 2) Package all 3rd party libs needed by *any* workunit into a single
zip with the main python library. These libraries would be installed into
the site-packages directory of the zip (i.e.
python25.zip/site-packages/NumPy) and the zip would be part of the science
app (again for cross-platform support).

It seems like option 1) is much more flexible and possibly more efficient.
However, option 2) is closer to the way that python normally works and
easier for the end user to understand and setup (by virtue of the fact that
they can simply zip their existing /Lib directory of their Python install).

I am leaning toward option 1), but are there any other possibilities or
suggestions? What seems most logical?

--
Jeremy
_______________________________________________
boinc_dev mailing list
[email protected]
http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.

Reply via email to