[directory-discuss] Fwd: Re: Debian/Ubuntu Database import

[Andrew Engelbrecht]

Hey, I sent this email almost a week ago, but I just got an email saying
it wasn't delivered. Here goes another try (after a few edits)... :)

-------- Original Message --------
Subject: Re: [directory-discuss] Debian/Ubuntu Database import
Date: Thu, 29 Mar 2012 22:53:15 -0400
From: Andrew Engelbrecht
To: address@hidden,  Michael Faille

On 03/26/2012 11:28 PM, Michael Faille wrote:
> As I understand, I think the goal is : Respecting the priority of the
> source of information for software : 1. official gnu package. 2. 
> triquel (up to date for others softwares) or anyother fsf distro 3. 
> other ?

I think this is a good idea. Karl, you're right, we shouldn't use
package names from debian or ubuntu. If there is any other database that
uses ubuntu package names, we can always map from Trisquel to ubuntu.
I'm guessing that wouldn't be much work compared to the rest.

As for version numbers, your're right Karl, they are more complex. But I
do think that they can be broken apart, at least by human eyes. I don't
know the exact rules, but sometimes they'll put a '1:' before the normal
version, and a '-4' or '-debian1.2-5' or whatever afterward.

> One concern, why we just not merge Trisquel directory, fsf directory
>  and X ?

I'm not totally against that. With that said, I think there are some
merits to the Semantic MediaWiki setup that we have, including its
hackability, and the fact that we've put a considerable amount of effort
into it so far. Getting data from Trisquel is a good idea, but I don't
think that Trisquel's package database could be a replacement by itself.
This is because there are many packages and much information in the
directory that is not in Trisquel.

> In the case of the GNU projects, we can duplicate some information 
> like last version and their date .

Yes. Karl, I couldn't find your script from the link you gave. How
complete is it at this point?

> Another suggestion : we can give  one "Stamp" on copy left project ( 
> protective licence) and another for GNU project.

I like that idea. I think that can be done automatically, if the
licenses are entered on each project page. (and many already have one or
more listed.) Some problems that need to be worked out are what to do if
there are multiple licenses for the project, and either all, or only
some of them are listed. Also, licenses change, so we might only want to
do it for recent projects.

> PS : integrating solr and mahout classifier could be useful for the 
> project Solr description :        http://lucene.apache.org/solr/ 
> Mahout description:    http://mahout.apache.org/ Example :
> 
> http://knackforge.com/blog/selvam/integrating-solr-and-mahout-classifier

Michael, would you please describe what your ideas are for how we might
integrate Solr and Mahout into this process? Since Semantic MediaWiki is
already semantic, we wouldn't need to run it server-side in order to
improve search results for users of the directory. But if we can use it
to somehow make the list that associates between directory page names
and Trisquel package names, to help us eventually import some data, I'd
like to hear your ideas about it. :)

Thanks!
-Andrew

> --- Michael
> 
> 
> On 03/26/2012 06:42 PM, Karl Berry wrote:
>> Hi Andrew, welcome to Michael, and ...
>> 
>> get updated project version info from some distro's repository data
>> and onto directory entry pages.
>> 
>> I really really really don't think that should be done for official
>> GNU packages!!  (Such as GIMP.)  Distro releases of GNU package XYZ
>> are, in general, not the same as the original GNU release of XYZ.
>> 
>> FWIW, I have accumulated a file of information about each and every
>> GNU package (and scripts to parse it) which I have always intended
>> to autoupdate into the Directory.  I don't think there is any
>> stopper at this point (there used to be many), but it has not come
>> to fruition.  It is the file gnumaint/gnupackages.txt in 
>> http://savannah.gnu.org/projects/womb if anyone cares to look.
>> 
>> planned database import from either Debian or Ubuntu.
>> 
>> As a separate issue, neither Ubuntu nor Debian are free distros 
>> that the FSF can list 
>> (http://www.gnu.org/distros/free-distros.html), so it seems a bit 
>> strange to be using them as the basis for a directory import, 
>> instead of (say) gNewSense or trisquel.
>> 
>> Best, k
> 
>