[Popcon-developers] Bug#707951: popularity-contest: please leak less private information

Bill Allombert Bill.Allombert at math.u-bordeaux1.fr
Sun May 19 21:21:20 UTC 2013


On Sun, May 19, 2013 at 09:44:24PM +0200, Bernhard R. Link wrote:
> * Bill Allombert <Bill.Allombert at math.u-bordeaux1.fr> [130512 12:39]:
> > On Sun, May 12, 2013 at 11:53:17AM +0200, Bernhard R. Link wrote:
> > > Package: popularity-contest
> > > Version: 1.57
> > >
> > > Please do not send second resolution information about program usage.
> > > Best only send out information what is actually used by the resulting
> > > graphs (i.e. a per-package NO-FILES/OLD/RECENT-CTIME/VOTE information
> > > and nothing else).
> >
> > I am considering rounding the number of second to the next multiple of 24h.

I have made a git branch 'bill-trunc_time' for that purpose.
Please check.

> > However, unless you are using strictatime, you probably do not leak much
> > information already.
> 
> Doesn't relatime update atime when it is older than a day? So doesn't
> relatime/strictatime just change from "second of last use before popcon
> run" so "second of first use in a 24 hours window, but still exact to
> a second"?

Not exactly. It is only updated if it is 24 hour old.
So you get the exact time of an event you cannot qualify. This is the stopped
watch fallacy. [1]

> > It is important the vote determination is done in a centralised way.
> 
> How does that prevent not sending timestamps?

If you do not have the timestamps, then you have to rely on the client
computation being correct.

> > > Additionally it would be nice to have a blacklist of packages to not
> > > send information from. Or perhaps some filter on packagename
> > > (mycompany-*) or sections (local/*).
> >
> > I am considering to allow packages to opt out of popcon by adding a control field
> > like "X-Popcon: no". Would that be suitable ?
> 
> That means you have to consider that when creating packages, which would
> be quite complicated to get retroactively.

If the package is leaked, it is too late anyway.
At least this would prevent the package to be leaked by forgetting to update
the blacklist on one computer.

Cheers,

[1] See Lewis Carrol complete work in your local library. 
-- 
Bill. <ballombe at debian.org>

Imagine a large red swirl here. 



More information about the Popcon-developers mailing list