Collecting information

[Rahul Sundaram]

Hi

Fedora needs a mechanism to collect user and hardware information with a 
opt-in mechanism. Similar lines of discussions here

http://mail.gnome.org/archives/desktop-devel-list/2005-July/msg00042.html

What do you think is the best way to do this?

regards
Rahul

[Matthew Miller]

On Tue, Jul 05, 2005 at 12:27:30PM +0530, Rahul Sundaram wrote:
> Fedora needs a mechanism to collect user and hardware information with a 
> opt-in mechanism. Similar lines of discussions here
> http://mail.gnome.org/archives/desktop-devel-list/2005-July/msg00042.html
> What do you think is the best way to do this?

Not sure about how to do it, but it'd be nice if it also included package
installation/use info a la the Debian Popularity Contest
http://popcon.debian.org/.

[Rahul Sundaram]

Matthew Miller wrote:
> On Tue, Jul 05, 2005 at 12:27:30PM +0530, Rahul Sundaram wrote:
> > Fedora needs a mechanism to collect user and hardware information with a 
> > opt-in mechanism. Similar lines of discussions here
> > http://mail.gnome.org/archives/desktop-devel-list/2005-July/msg00042.html
> > What do you think is the best way to do this?
>
> Not sure about how to do it, but it'd be nice if it also included package
> installation/use info a la the Debian Popularity Contest
> http://popcon.debian.org/.

Ok so the following tasks then

* A tool to collect package stats
* A tool to collect hardware information
* A form to fill out for users to send in more information to the 
project with appropriate "incentives" to do it

regards
Rahul

[Alexander Dalloz]

Am Di, den 05.07.2005 schrieb Rahul Sundaram um 14:27:
> > > Fedora needs a mechanism to collect user and hardware information with a 
> > > opt-in mechanism. Similar lines of discussions here
> > > http://mail.gnome.org/archives/desktop-devel-list/2005-July/msg00042.html

I like the idea to know something about the user base (how much users in
the different countries, gender proportion [Fedora attractive for female
users?], Gnome | KDE rating).
> > > What do you think is the best way to do this?

I feel it is a must that the data collection is on a volunteers basis.
It must be easy to understand, easy to give away data - to know to whom
and how much personalized! - and easy to quit submitting the own data.
Clear list of transfered information.
Any kind of doubts like with the Microsoft Windows® build-in update
process must be strictly avoided.
> > Not sure about how to do it, but it'd be nice if it also included package
> > installation/use info a la the Debian Popularity Contest
> > http://popcon.debian.org/.

Nice too. Could be an indicator what to have in Core and what in Extras.
Rating third party repository popularity.
> Ok so the following tasks then
> A tool to collect package stats
> A tool to collect hardware information

Remembering back the days of Red Hat Linux up2date did something like
that. So some code should be there (please don't ask me how much it
qualifies for Fedora purposes). RHN registration was an enforcement back
in these days to be able to use up2date.
> A form to fill out for users to send in more information to the
>
> project with appropriate "incentives" to do it

Yes, must be attractive for the Fedora user to spent time and to offer
data. Fedora project gains data, the individual user feels bound to the
OS in some way ("proud to be Fedora")
> Rahul

Some unsorted quick thoughts:
- reliable insurance that data can't be misused and be removed by user
on request
- menu entry or applet for ease of use
- applet showing the current Fedora population
- world map with locations of registered users (not naming them, not
required form entry)
- each registered user gets his personal registration number (like
http://counter.li.org/), maybe some kind of "certificate" like the
seti@home project generates one for number of processed seti packages;
counter for age of Fedora community "membership"?
- heartbeat functionality to release old database host entries where
user didn't signed off
- once done initial data registration user must get explicit note about
scheduled recollection and must be able to cancel processing

Alexander

[Jef Spaleta]

On 7/5/05, Rahul Sundaram  wrote:
> A tool to collect package stats
> A tool to collect hardware information
> A form to fill out for users to send in more information to the
>
> project with appropriate "incentives" to do it

I'd rather know why we..as fedora..are collecting any of this
information before we collect it.
I want to see concrete plans on how this information is going to be
used before we start collecting.  Who is going to datamine that
package stats information and for what constructive purpose?  Who is
going to datamine that hardware information and for what constructive
purpose?  Who is going to datamine geographic and other userdata and
for what constructive purpose?

I am very much not inclined to building a treasure trove of data...
just to have it around because people are somewhat interested in
knowing things about the userbase.  If there are no concrete plans to
use the data as it comes in, I think having a mountain of data its
only going to prove to be a burden.

Gnome has a stated goal cleverly named "10x10". Getting an estimate of
the number of gnome users is clearly needed to know if they are on
track to reach that specific goal.  Fedora on the other hand, has no
grandiose user adoption targets. I don't see a driving reason to build
this mountain of information up.

I don't see any credible discussion that people agree that the popcon
concept is something Fedora is willing to adopt  or any other need for
use package stats data. I don't see any credible discussion about user
adoption targets that could possible need userdata for feedback. I
don't see any credible discussion about doing something worthwhile
with hardware data if we had it.

All i see are some vague expressions of desire to leisurely mine data
to get some trending statistics. I don't think that is a good enough
reason to start building up the data. I think there needs to be
commitment to actually use and process the data as it rolls in.
Sitting on a mountain of userdata without doing anything with it as it
comes in, is just going to lead to a grotesque mess as people with
specific agendas start mining historic data to try to strong arm
discussion.

-jef

[Elliot Lee]

On Tue, 5 Jul 2005, Rahul Sundaram wrote:
> Fedora needs a mechanism to collect user and hardware information with a 
> opt-in mechanism. Similar lines of discussions here
>
> http://mail.gnome.org/archives/desktop-devel-list/2005-July/msg00042.html
>
> What do you think is the best way to do this?

The 'install coverage testing' project would have done this. I have a
really good idea how it 'should' work. The main challenge is finding
someone with time to work on it.

Best,
-- Elliot
You can accomplish anything you want, so long as you don't care who gets credit for it.

[Rahul Sundaram]

Jeff Spaleta wrote:
> On 7/5/05, Rahul Sundaram  wrote:
> > A tool to collect package stats
> > A tool to collect hardware information
> > A form to fill out for users to send in more information to the
> >
> > project with appropriate "incentives" to do it
>
> I'd rather know why we..as fedora..are collecting any of this
> information before we collect it.
> I want to see concrete plans on how this information is going to be
> used before we start collecting.  Who is going to datamine that
> package stats information and for what constructive purpose?  Who is
> going to datamine that hardware information and for what constructive
> purpose?  Who is going to datamine geographic and other userdata and
> for what constructive purpose?
>
> I am very much not inclined to building a treasure trove of data...
> just to have it around because people are somewhat interested in
> knowing things about the userbase.  If there are no concrete plans to
> use the data as it comes in, I think having a mountain of data its
> only going to prove to be a burden.

Marketing: Where do we stand in terms of number of users and what kind 
of users Fedora is attracting
Development : What type of hardware works,  Whats packages is being 
utilised more?.  What can be moved into extras?

regards
Rahul

[Jef Spaleta]

On 7/7/05, Rahul Sundaram  wrote:
> Marketing: Where do we stand in terms of number of users and what kind
> of users Fedora is attracting

You need to be very careful here...  Are there plans to attract or
discourage a specific "kind" of user? I'm not sure knowing precise
numbers as to the demographics of users provides any useful
information unless there is a stated goal or a commitment to service a
specific demographic or a stated goal or commitment to reach certain
adoption levels (relative or absolute) among different demographics. I
haven't seen any discussion about aims or goals about demographics
targets. If Fedora installs seem to be relatively low in Finland
compared to Brazil, does it matter?  Is there a number of installed
systems target we are shooting for?
If Fedora is popular with one "kind" of users and not so popular with
another "kind" of user... do you work to balance that, even if it
means making it less popular with the larger group? Or do you aim
efforts at making things more appealing to the "kind" of user already
drawn to the project and the distro?  I think these larger questions
need to be answered before you collect any demographics data.  As I
said before, Gnome at least has some (wacky) stated adoption goals for
which that want data for to use to track progress towards that goal.
Fedora doesn't have any adoption goals really, and until we agree on
goals with regard to adoption demographics I don't see a reason to
collect data about it.
> Development : What type of hardware works,

Is there a commitment from anyone to actually mine that data as it comes in?
And just as importantly... to follow up on reports of brokenness as
updated packages come.
A usable HCL list is going to take real manpower.. and can't be
automated. Creating a big data dump without someone committed to
actively maintain it seems counter-productive to me.  You run the very
real risk of having issues linger as "broken" well after they are
fixed.   if it takes 6 months for a human being to actual mine data
about broken hardware... thats not particularly useful information for
this project, thats a pretty long time scale in comparison to how
quickly update packages get pushed out that impact hardware.  This
might be useful raw data for developers,  I'd like to see what the
kernel and X.org developers think about auto-collecting data about
"working" or "broken" hardware and if having a big datadump like that
would be useful for them in terms of identifying problem areas. If
they don't find it useful to build up databank of "raw" data, its
probably not going to be worthwhile for end-users.  An end-user
oriented HCL is going to require a team of people who are committed to
sifting through the reports and following up on broken issues as they
get fixed with updates quickly.
I want to see people step up and commit to that before data is collected.
> Whats packages is being utilised more?.  What can be moved into extras?

Even if you could measure the utility of a package in an accurate way
that reasonable accounted for the skew related to "default
installs.".. i've seen NO credible discussion among the release team
nor the Core developers that popularity or utility is an important
criteria for any decisions with regard to Core.   I do not want data
to be collected about popularity until there is consensous that
relative "popularity", "installed", or "usage" of a package is
something that is actually going to impact decision-making by the
people who actually make those decisions.  Taking this data, without a
commitment to use this data.. is just going to create opportunities
for people to argue unconstructively.  First get agreement from Core
release team and maintiainers that "utility" or "popularity" of a
package is important to the decision to keep it in Core or not, then
take the data if that is the concensous view of the people who make
these decisions.

-jef