Hi all - this is a topic near and dear to my heart so please forgive
the verbose answer.|
Up front I should say that I think personas are a key component of good UX, and without some kind of user definition you're going to run into problems, including the elastic user problem. Which typically can lead to scope creep, and project delays, and terrible, terrible things. Personas can help - in the right situation.
I think personas are a very "modern" UX deliverable, but they're not well used in industry, to be honest, and even less so in Open Source projects (strictly from my experience so far). Many development shops claim to want them, and then aren't sure what to do with them once they have them. All they really wanted was a better understanding of their users and their needs, and they aren't sure how personas gets them that once they have them in their hands. This problem is compounded by the fact that not everyone agrees on what personas are, nor what they should contain.
If we're creating generic Gnome personas, how will they be used? Are we looking for a tool that helps us publicize and market Gnome? If they're intended to capture end user behaviour and needs - which Gnome project will be using them? If we create them and nobody uses them, is that OK? If someone uses them, we really should make sure they're as useful as possible to those people. If we make them too generic then they will simply be ignored as being unhelpful and unspecific.
We can consider the existing personas that have been drafted/created in various Gnome projects [1, 2, 3, 4] and ask how we can improve upon them (or even use them). Or look at some existing user observations  and ask how they were used, and how to use them more.
These are just some of the thoughts popping into my head.
In this case it looks like there is a desire to better understand the general Gnome user population. I'd suggest trying to focus the personas on a market approach, rather than a software tool approach, as it seems like they'll most likely be used as marketing tools. Or perhaps take this opportunity to collect data on users and present it in a data-rich way, rather than as stereotypical users.
If we're able to do a large dump of user data I'd love to look at some ideas I was throwing around back in the early days of Gnome Shell  - taking a look at existing Gnome desktops, how they're configured, and how they're modified from the default setup There's a link on that page to a presentation done at MIX09 by Microsoft  on how they gathered data to retool the Windows taskbar for Windows 7. I think having this kind of data would be greatly valuable for designers and usability analysts alike in trying to figure out how to improve the user's experience.
Although on second thought the panels all go away with the next version of Gnome, so this might require more thought in terms of what exactly we want to gather data on.
 Banter personas: http://live.gnome.org/Banter/Design/Personas
 Orca personas: http://live.gnome.org/Orca/Specification/Personas
 Gnome web personas: http://live.gnome.org/GnomeWeb/WebPersonas
 UTS personas: http://live.gnome.org/UTSPersonas
 GnomeShell user observations: http://live.gnome.org/GnomeShell/UserObservationData
 Gnome Shell user analysis ideas: http://live.gnome.org/GnomeShell/UserResearch
 Windows 7 user investigation video: http://videos.visitmix.com/MIX09/C26F (Silverlight required of course (boo!))
On 10/30/2010 09:29 PM, Allan Caeg wrote:
I'll add some more things to ask based on what I learned from my profession.