Contents
1. Purpose
This page is for discussion (an maybe a description of implementation!) of the best way to count the number of GNOME users. This is needed on order to assess progress to the 10x10 goal.
Discuss these ideas real-time on the #marketing channel on irc.gnome.org
Why do we want to do this?
So we can:
Measure progress toward 10x10,
- Have (motivational) feedback for advocates,
- Enable potential users to make better investment decisions (time and money) by tracking momentum,
- Include hard data for press releases to accelerate/encourage the adoption process.
2. Ideas
Ideas for counting users discussed here should bear in mind the following points:
- Counting downloads is problematic because downloaded ISOs may be copied many times
- Many GNOME users are using network booted terminals
- Counting Linux/BSD/Whatever installations is not so good because many are used as servers only
- It seems that the GNOME Foundation Board, Red Hat, Novell etc. do not have these numbers either (can anyone confirm that?)
- We want to catch current users, so any method will have to account for switching to and from GNOME
- Distributions can, and do, alter/customise GNOME for their releases. Thus we must implement any client-side solution in such a fashion that distributions have no reason to alter it
There are at least two ways to achieve more valid counts than are currently available:
- Examine server logs for hits; or
- Include software in GNOME that "phones home".
2.1. Server Side
Types of server that have been discussed are *.gnome.org, TCP and gnome-weather servers. However the problem with this approach is that it may not catch all (or a reasonable proportion of) users. Of the options considered, TCP servers will probably catch the most users, but those servers are typically (or all?) controlled by distribution vendors. This makes us dependent on the goodwill of these vendors, who may not wish to share that information. What's in it for them?
2.2. Client Side
Including in GNOME some piece of software that contacts a central GNOME server has also been considered. The problem with this is that if it requires user effort then users are less likely to use it. And the less user effort it requires the more ethical issues are raised. The limit is a daemon or script that the user is totally unaware of; this would be unethical to some (most?) people.
2.3. Currently Preferred Option
Currently the best solution that has been discussed it to modify the "About GNOME" dialog box include a link to register your GNOME installation/user-ship/love/whatever, or alternatively a toggle or check box that indicates the user's permission to activate a service.
3. Details
Any information should include:
a hash of username (?) (We want to measure users, not installations) (ClausSchwarm: I think, we need both: users and installations!)
- GNOME version
- Timezone/country code (measure penetration/share on a country-by-country basis)
- The software should report periodically, to address the switching/currency issue.
- The reported information should be made visible to the user on request.
- The user (or the admin?) can enable the software
This information is for us. However if we included information on the distribution that GNOME in running on (and perhaps some demographic details of users?) this would be commercially valuable information. We could either share this with distributions in a show of goodwill; or sell it to them in order to fund worthwhile causes. (Further research, scholarships, ...)
4. Implementations
None yet. We like the idea of a toggle button (Suggested by DavidMorenoGarza) best, but probably disabled by default.
5. Additional needs
- A 'phone-home' feature will certainly grasp the attention of the press. Thus, we must deal with it proactively, that is: announce the inclusion of such a thing before the release notes, with proper explanations why we need to do this, in the best of all worlds immediately after a decision.
6. Comments
DavidMorenoGarza: Alternatively we could use a toggle (enabled by default) button there. That way, unexperienced users could also never realize of its existence, but they would have the choice of turning it off. If any ethical objections are raised, we will have the results available for everybody.
JohnWilliams: Could you explain or re-write the last sentence please? It is precisely the point that having software automatically expose information about the user and/or their computer without the user's knowledge that is not ethical. How would having public information available assuage these concerns? And how could a user have the choice of turning the toggle off if they are unaware of its existence?
SteveFeehan: How about a one time registeration that results in an email alias (my.name@users.gnome.org), jabber account, etc. Then ongoing use could be tracked by use of the account rather than a program "phoning home". This account would give the user a feeling of ongoing "connection" to the Gnome project, could be transported to other computers (tracking users not installations), tied to wiki/bugzilla account, etc.
JohnWilliams: Why would a user do this if they already have an email address/account? I know some would, for the reason you mentioned, however the point of this exercise is to count as great a proportion of users as possible. Also, it would rely on people sending mail to that address, so two levels of choice reduce the chances of success hugely.
ThiloPfennig: Does this help: http://popcon.debian.org/ ? It tells how many of all reported installations have installed GNOME on Debian.
JohnWilliams: popcon has been discussed and rejected because it must be installed and activated by the user.
ClausSchwarm: It would nevertheless been helpful if somebody could analyze the numbers. They were not collected from a 'Which desktop do you prefer?' point of view and any results could thus serve as a counterweight to the various web polls. Any takers? Thilo?
ThiloPfennig: I am not a statician. I could collect some raw data that I think is helpful and we could then tell how many of those who helped creating this data actually used GNOME. This needs to be: not only core but a runable GNOME (package gnome-session?).
JohnWilliams: I am a statistician. Give me the numbers and I will do what I can. But remember: GIGO
JohnWilliams: Following from Steve's idea above, perhaps simply a web-form on www.gnome.org for people to fill in as much or as little information as possible, with the least being that they use GNOME and their email address. Then we could email them every year or six months to confirm they are still using. Have it so that simply replying to the email could be taken as "yes". This would probably not catch a large portion of users, but it is something we could do right now, with minimal coding effort.
- ThiloPFennig: or the get an email adress and can confirm via URL, like mailing lists do. the also could get an email if a new version of GNOME is out. If they also tell us how they use GNOME we could also let them know how and where they could update, like "Fedora Core will have GNOME 2.12 in its next release that is scheduled for February 13th. ..."
AlanHorkan: There already exists a Feedback tool in Gnome and it is called Bug Buddy. The information Bug Buddy gathers could be expaned to include more information including information suitable for Marketing purposes. The Netscape Talkback tool provides an example of how to do this and it also clearly provides a way for users to decline to send certain information they would rather not share. This could probably be done in addition to any other plans you might have.
JohnWilliams: This idea has been discussed and rejected, mainly based on the fact that any distro would be likely to modify bug-buddy to remove this feature. I have no clue on this: I am going on Jeff Waugh's advice here
* MathiasHasselmann: If you want to count users by phoning home, without getting ethnical problems, this home call must have some benifit for the user allowing his GNOME to call home. Examples are:
- checking for updates
- fetching a useful news-letter/rss-feed
- moving parts of GNOME's help system in the web: default installation contains just the basics, annotations, updates, ... happens in the web
- searching, spell checking, dictionary: also could use some web service at gnome.org
- you get the idea
* PierreSlamich What about a 'Tell the world i'm using Gnome' on first startup.Could be customised by vendors into 'Tell the world i'm using Thingie Linux' but Gnome would be pinged as well.
* ghaywood: What if there is a short user registration (single box) maybe the user details box you get under "Preferences/user details" when a gnome session is created for the first time, just asking for some details, with an option "I am a new user and I would like to be counted" sort of thing?
* LloydAlmeida : What if gnome.org could provide a some space for users in the domain gnome.org, say a couple of megs that store gnome settings ( eg. theme, fonts, wallpaper etc.). This would solve two issues one for users wanting to migrate their preferences to multiple machines (eg. laptop to desktop etc. ) via a multiple profile kind of setup and two a way for GNOME to count exactly how many machines and users are using it ).If this was implemented in GNOME as a feature it would be in all distros.
7. Results
None yet. (Aaargh!)
8. Resources
Put links here that relate to estimates of
The total number of computer users worldwide
- The total number of Internet users worldwide
- The market share of operating systems
- The market share of Desktop environments that work any of the operating systems that GNOME works on
An account of an attempt to estimate of the number of Internet users: http://www.blogcadre.com/blog/jason_striegel/weighing_the_internet_2005_07_13_03_37_07
Another estimate (probably more valid): http://www.internetworldstats.com/stats.htm
An estimate of OS share by Internet traffic: http://www.w3schools.com/browsers/browsers_stats.asp
Desktop market share of GNOME (2003-2004) from a poll at Desktoplinux.com: http://www.desktoplinux.com/articles/AT2127420238.html