--- Begin Message ---
- To: submit@bugs.debian.org
- Subject: popularity-contest: optionally gather machine details
- From: Matt Taggart <taggart@debian.org>
- Date: Sun, 13 Nov 2005 23:53:26 -0800
- Message-id: <20051114075326.5C1DF1AA8BB@cyrix.home.bogus>
Package: popularity-contest
Version: 1.31
Severity: wishlist
It would be nice if popularity contest could optionally gather information
about the computer it's running on. Here are the things I've thought of:
* What vendors and models of computers run Debian.
This might be useful for convincing a vendor to help support Debian on their
machines. So many vendors equate what what O/S a machine is running with what
O/S it was sold with, when in reality we all know people install Debian over
the top of whatever the machine came with once they get it :) These details
could be gathered with debconf questions. (Disclaimer: I work for HP and I
want this to be able to convince my management what systems to support)
* What amounts of memory and disk space a machine has.
This would allow us to determine what resources the installer, packaging
system, and other software have to work with. After asking permission with a
debconf question, this info could be gathered from /proc.
* What I/O devices a machine has.
This would be useful for the kernel team to determine what drivers to support
and how. After asking permission with a debconf question, this could be
determined with lspci/lsusb/dmidecode.
* Where on the network in relation to Debian this machine is.
This would allow us to determine where mirrors are needed. We'd have to be
careful on this one to not reveal too much in the info gathered where the
machine is. I'm thinking gathering the number of traceroute hops to several
mirrors or something like that.
* What city and country the machine is located in.
This could be used for cool maps and graphs and to lobby governments to
support Debian. Gathered by debconf questions.
* Number of users of the machine.
Interesting. Might have practical uses for default daemon settings or package
tuning.
Probably bucketed somehow, like 1, 2-10, 10-20, 20-50, 50-100, 100-500, 500+.
I guess it depends on how you define users, you can have a lot of 'users' of a
website. This will need some discussion.
* What the system is used for.
Helpful in determining stuff for tasksel.
Checkboxes of various things people use debian for, maybe including an "other"
where they can put in their own use. (if particular 'other' options show up a
lot they could be added in future releases).
* What the system is primarily used for.
Sorting for tasksel, other uses.
radio buttons based on answers to the previous question
* class of system.
Interesting. Maybe useful for lobbying industry groups.
Radio buttons of something like 'embedded device', 'desktop', 'server',
'mainframe', 'kiosk', 'cluster'. These selections will need some discussion.
'cluster' in particular might be neat because it would allow us to determine
if clusters were artificially inflating other popcon numbers (a claim I've
heard).
Again all of these would be optional.
It might be interesting to store the results in rfc822 format somewhere on the
system, maybe /etc/system or something? A standard for that would be neat. We
wouldn't want that abused though, need to think about that.
Thanks,
--
Matt Taggart
taggart@debian.org
--- End Message ---