[Dev] package popularity tracking (Was: repo redirector modificatoins)

Andreas Grapentin andreas at grapentin.org
Tue Jun 11 19:39:24 GMT 2019

On Tue, Jun 11, 2019 at 03:05:12PM -0400, bill-auger wrote:
> sure, but that would be either voluntary or an extra requirement
> to place on mirrors - we dont need a total sample set though - i
> think most users do use the redirector, as it is the only source
> enabled in default mirrorlist - if so, it should be able capture
> statistically significant results, without requiring or imposing
> anything on mirrors - we have so little information about usage
> now, it would be difficult to determine even that much

To be fair, the first thing I do on a freshly installed parabola system
is change the mirrorlist. simply because the repomirror used te be so

That being said, I don't think we need to track downloads on all
mirrors. Assuming statistically equal spread of requests, tracking
successful downloads on *any* mirror would do, at least to determine the
relative popularity of packages against each other. Tracking the
repomirrors redirects would be useful to get a measure on the overall
popularity of the distro, i.e. number of installs that actually use the
repomirror, but I'm not sure what that measure would be good for.

I propose choosing one or two reliable and well-used mirror we have
direct control over, and track successful downloads there, and
extrapolate the results. Let the redirector do its thing.

> it is true though, that any user who prioritizes a specific
> mirror but still keeps the redirector enabled could skew the
> results - that is the recommended configuration in the mirrorlist
> file: either to use the redirector exclusively, or to keep it enabled
> with a lower priority to a preferred mirror - maybe that
> recommendation would need to be reversed to suggest the
> redirector be used only mutually exclusive to any other explicit
> mirror, in order to gather representative samples

Putting the redirector at the end of the mirrorlist would make a lot of
sense. It would allow users to enable (uncomment) a default mirror they
prefer, and pacman would only fall back on the redirector if needed. If
we want to continue this train of thought, I would recommend doing so in
a new thread :)



my GPG Public Key:                 https://files.grapentin.org/.gpg/public.key
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 488 bytes
Desc: not available
URL: <https://lists.parabola.nu/pipermail/dev/attachments/20190611/c4e1bd04/attachment.sig>

More information about the Dev mailing list