Re: [tdf-discuss] Better defaults

On 27/10/10 20:41, Marc Paré wrote:
> Le 2010-10-27 15:30, AG a écrit :
>
> > Marc
> >
> > I had in mind some kind of sub-routine that, with the user's sign in,
> > would use system data (language preference, locality) along with LibO
> > data (which components are used and relative frequency) and periodically
> > send this data in an anonymised (scrubbed) manner to the LibO team for
> > statistical analysis.
> >
> > The process would have to be transparent, open to user scrutiny and
> > something that they select to do as a contribution to the overall
> > project development, much in the way that SETI would use spare cycles to
> > crunch numbers and send off, in a similar way LibO could request that
> > users sign up to send preference stats.
> >
> > I'm no programmer so probably am not using the appropriate jargon, so I
> > hope this clumsy description outlines the idea. The advantages are: the
> > user doesn't have to register nor do anything aside from tick a box
> > during LibO installation and second, it does lay the ground for
> > promoting LibO as a community-led concept that has a "built in" means of
> > recruiting and then using user feedback. The latter available through
> > selection - like a continuum of data to send back to LibO including
> > explicit feedback opportunities and joining a community to pool
> > experience and expertise.
> >
> > Something like that, anyway.
> >
> > AG
>
> I don't think I would favour this. There is too much potential for 
> gathering data from users and we could really run in trouble if 
> criticism were to spread that we are data mining our LibO users.


Well, that's why it is all open and above board, the code is open, the 
fields are explicitly defined, clear data specifications, and the system 
mails out to LibO (cc the user) the reports.  It isn't data mining: it 
is voluntary, the default option is no, it can be stopped at any time, 
it is participatory to the levels of detail the user is okay with, it 
invites open transparent scrutiny, and posts the user a copy of the 
report being sent.

Moreover, it wouldn't affect the installation of LibO - it is a 
voluntary participation in research and users can opt in if they want, 
and if not, then fine.

I'm not defending the idea, nor am I saying that this is what 
"popularity contest" does.  I was asking after how feasible a basic 
approach like that would be for LibO community's purposes.

> Is there anything like that on the debian project?
>
> Marc
Here's the scoop on the popularity contest itself, launched early 2004:

"The Debian popularity-contest is a concept created by Avery Pennarun a
few years ago.  It set up a program on the hosts installing the
popularity-contest package, to email the list of packages installed
and in use to a central collection point.  It also collect the host
architecture and we plan to collect kernel version and modules used as
well.  The summaries are presented on <URL:http://popcon.debian.org/>
and used to sort the packages on the Debian CDs.

The information can be used in other areas as well.  It can detect
which packages in the debian archive which aren't installed on any
hosts.  Such packages should probably be checked out and possibly be
removed from the archive.  It has already been used to check which
non-free packages are actually in use, while discussing the future for
non-free."

http://lwn.net/Articles/75753/

AG

--
Unsubscribe instructions: Email to discuss+help@documentfoundation.org
Posting guidelines: http://netmeister.org/news/learn2quote.html
Archive: http://www.documentfoundation.org/lists/discuss/
*** All posts to this list are publicly archived ***