We would like to fix up the source and data repositories used by NIPY to provide an easier way to add sample data sets, while reducing the amount of data a user needs to download to get NIPY up and running.

Aims

• The basic download of NIPY should be minimal. Without any example data, it should be possible to install with a download of <10MB (and probably <5MB if we can get enthought out of the tree). JM: I think that there should be some data included with any svn download. If you are using svn, you should be running tests. So I think we need to figure out what the minimal data set needed to run the basic tests. Installing NIPY for end-users shouldn't require data; but endusers will install a binary package and won't use svn.
• Extra data should be optionally available for svn download if wanted. We should also make some data packages for endusers as well.
• Unit tests/examples which depend on this extra data should prompt the user to download the extra packages if they aren't found.

Proposal

Create a "NIPY data pack", which is a place for keeping all large pieces of data which might be used for either testing or examples. These would live in their own separate svn tree and not be required for installation and basic testing of NIPY.