[Popcon-developers] popcon & derivatives

[Bill Allombert]

On Mon, Jan 20, 2014 at 09:48:09AM +0800, Paul Wise wrote:
> Ah, the whole idea behind including dpkg vendor information was to
> make sure we could tell derivatives users from Debian users so we
> could ensure derivative users do not influence the Debian statistics.
> For the initial stage I assumed reports from derivatives users would
> be saved but be skipped during report processing, I guess that isn't
> the case yet?

No, they are treated identically to other reports. Personnaly, I did not plan
to do otherwise.

> I figured later down the track we could work out how to
> include derivatives in the processing and how to display them
> separately on the website.

As I said this is theoretically possible but not cost free.

> The reason I suggest the version number stuff is that not all
> derivatives are using the latest version of popcon and not all of them
> will be setting the dpkg vendor correctly. Falling back to detecting
> via the popcon version number will catch some users (admittedly few
> right now).

Until this raises to the level of a practical issue (instead of concerning
half a dozen submission as of now) there is no need to do anything.

> If you wanted to do processing of reports from derivatives, the
> derivatives census includes sources.list snippets and daily downloads
> apt Packages/Sources for each of the derivatives we have in the wiki.
> I think the census includes all of the vendors I've seen you mention
> here as well as the ones listed with popcon version number extensions
> on popcon.d.o.

First, I think it is dangerous for privacy reasons to publish statistics about a
very small set of submissions.

Second, there is the question whether the users of derivatives are willing to
submit the data to Debian, which they might perceive as a third-party between
them and the authors of the derivatives.

Third, the cost of processing a distribution is proportional to the size
of the Package files plus the number of submissions. So even a 
distribution with a few reports requires some processing time.
(Now the popcon system could work differently, but that is a different story.)

So this is a significant undertaking which will use some Debian server
resources.  Whether it is worth doing it is an open question, and I do not have
time to work on this.

Are you interested by some particular statistics ?

Cheers,
Bill