[directory-discuss] Debian/Ubuntu Database import

[Andrew Engelbrecht]

Hello Michael and directory-discuss,

I'd like to introduce Michael Faille, who I met at the LibrePlanet 
conference yesterday. He said that he would like to help out with the 
directory, including the planned database import from either Debian or 
Ubuntu. It's a big job, and I know that Joshua has been discussing it 
with some people regarding how it should be done, so I hope he'll weigh 
in on our strategy.

For those of you who don't know about the plan, the first long-term goal 
I see is to get updated project version info from some distro's 
repository data and onto directory entry pages. Reading values about 
each package shouldn't be the hard part, but I think we will have the 
biggest challenge in matching up directory.fsf.org project urls/page 
names with a distribution's package names, since there will be some 
variation between the two lists. So there will have to be some manual 
matching and verification wherever it is challenging for an automatic 
script to find matches.

For instance, in the directory, the project name "GIMP" corresponds to 
"gimp" in the debian repository. That's an easy match for a script to 
find. However matching "Armadillo: C++ library" and "libarmadillo2" is a 
bit harder, and I think there are some that will be more challenging 
than that. One strategy for this issue could be to try auto-generating a 
list of possible matches, based on similar names and project homepage 
urls for each project in the directory. We could then to split between 
many people the task of human selection/verification.

And once that is done, we can write a simple python script using the 
mediawiki extension to auto-edit the "templates" on each project page in 
order to include an entry that lists the distro's package name. Then it 
will be easy to broaden the scope of data to import, such as updated 
descriptions, since the groundwork will be laid. For instance, Joshua 
was telling me and Michael that there is another database that lists 
extra information beyond what's in bare-bone debian "Packages" files. 
This info is referenced by debian or ubuntu package name.

So Michael, while I was a bit unclear in my original description to you 
at LP, I hope this gives you a better idea of what we have before us. If 
you need a better explanation from me, I can answer questions. If this 
is indeed something you wish to help us with, we would all love to hear 
your thoughts.

Thanks, and welcome aboard. :)


-Andrew