Re: [Banshee-List] anonymous usage data

[David Nielsen <gnomeuser gmail com>]

2010/2/12 Gabriel Burt <gabriel burt gmail com>

On Fri, Feb 12, 2010 at 10:54 AM, David Nielsen <gnomeuser gmail com> wrote:
>
>
> 2010/2/12 Gabriel Burt <gabriel burt gmail com>
>>
>> On Fri, Feb 12, 2010 at 12:16 AM, Alexander Kojevnikov
>> <alexander kojevnikov com> wrote:
>> > On 12 February 2010 19:11, David Nielsen <gnomeuser gmail com> wrote:
>> >> If one was to run with --debug-sql could we also report frequent long
>> >> sql
>> >> queries perhaps. I've noticed a lot of users complaining about
>> >> Banshee's
>> >> performance and while digging through logs I've noticed that it is
>> >> possible
>> >> to get queries lasting several seconds. It would be nice to see just
>> >> how
>> >> frequent this happens to users.
>> >
>> > I was sceptical about the whole idea until I saw this. I could
>> > definitely make use of this information.
>>
>> Yeah, it's a good one.  We can do it without requiring --debug, too.
>>
>> I have a rough first pass at this whole usage data/metrics work in
>> master now.  You can opt-in to it via Preferences.  It doesn't
>> actually send the data anywhere yet, but it will collect it.  So far I
>> have it collecting info on:
>>
>> * What assemblies are loaded/what versions
>> * Basic distro/os/.NET version info
>> * How long/frequently Banshee is run
>> * What source is active (and when it changes)
>> * What CultureInfo/locale we're in
>> * Track counts for libraries, podcasts, radio, etc
>> * Total play count / skip count
>> * How many tracks are rated
>> * Timestamp of when the app starts, exits (and duration)
>>
>> You can see an example Json dump of the collected data here:
>> http://banshee-project.org/~gburt/tmp/usage-data.txt
>>
>> Anybody should feel free to help out adding other interesting metrics!
>
> How about collecting data on which columns are enabled in which views and
> the order in which they are placed. To help us spot where the users
> frequently alter the Banshee defaults and thus indicating that we are doing
> it wrong(tm).

Great idea, I'm adding it to my personal todo list.

I forgot to add, we should probably also collect the browser setting (top vs. left), if we can justify it it might also be nice to know the screen data to that we can get an idea of which sizes we should optimize for. E.g. standard Banshee does extremely poorly on my netbook and the long awaited Cubano frontend hasn't yet arrived. I could also imagine that this could be used to figure out how many tracks we should load additionally to those displayed (my understanding is that we currently do something like displayed +/- 10 tracks).

We could perhaps also gather some data on the database itself to see how many revisions it has gone through, I suspect this would reveal any potential problems where we are not migrating correctly from none linear versions upgrades or from things like the daily ppa snapshots. This could be useful for distribution upgrades to ensure that the shipped version is known to upgrade nicely when the distro issues major upgrades.

Another nutty idea would be to collect the theme being used so that we might test that Banshee looks as it is expected to as part of the release process.

Finally we could collect data on which devices are routinely connected to Banshee, their storage capacity, transfer protocol. This would let us see which devices are most important to hit during release testing and a sudden dip in these metrics would also indicate that something broke. Likewise we should collect some data on the errors we see transferring to such devices.

- David