Wineconf follow up: Wine Usage Data Collection

[Austin English]

On Wed, Oct 8, 2008 at 2:34 PM, Chris Ahrendt <celticht32 at aol.com> wrote:
> Juan Lang wrote:
>>>
>>> I think we do Juan..
>>> reason being is that information is not always captured when people put
>>> things into bugzilla..  if we had a way that its entered initially (if we
>>> could gather it automagically)  then have them attach the file  when they
>>> do
>>> a bugzilla that might work a little better...
>>
>> I disagree that this should be a goal.  Our intention was to sample
>> the applications that people are running, and not to require it.  That
>> is, it should be an opt-in survey.  Getting better bugzilla reports
>> should be orthogonal to this, and would be a task better handled as a
>> crash tracker like Dr. Watson.
>>
>> And to preempt comments like, "Why not do both?", we shouldn't
>> provider better support to people that opt-in to our optional survey
>> than to those who don't.
>> --Juan
>
> so why not collect it period when wine starts up... dump it to a file and
> let people know we are doing it. Have the ability to turn it off..
> once the file gets to a certain size send it in if there is an active
> internet connection or zip and mail it off...
>
> chris
>

If someone only uses Wine periodically/infrequently, we're unlikely to see it.

Currently, we've got two problems:
1) We want to collect Wine usage data, so we know where to concentrate
our efforts.
2) We don't want to bug developers with this, because we'll frequently
go through dozens of wineboots in a day.

Current proposals:
1) Run a dialog box on first wineboot asking to opt in.
2) Disable the dialog box by default, but allow distros to enable it,
so we don't bother devs.
3) Run the dialog box periodically, reminding users about the survey.
4) Put an option in Winecfg to allow opting in, similar to Ubuntu's
installer's advanced options allowing to opt-in to the package survey.

Problems for propsals:
1) Will annoy developers, which means, won't happen.
2) Possibly viable, but we're then depending on the distros to enable
it. While Ubuntu/Suse probably won't be a problem (nudges
Scott/Marcus), the others may be.
3) More trouble to implement, and annoying.
4) Depends on users finding the survey and opting in. Will give
slightly biased info, as well as less than options 1 or 2 would.

I'm leaning toward 2 or 4. The others are annoying/not worth it.
Remember, we're not going to get EVERYONE. The idea, however, is to at
least get SOMETHING, so we know where/what to target. Currently, we're
shooting in the dark.

-Austin