[Popcon-developers] Please review new description and question text

[Bill Allombert]

On Sun, Jul 30, 2006 at 12:11:45AM +0200, Petter Reinholdtsen wrote:
> [Bill Allombert]
> >>  * Check files in ^/boot/ as well, to detect if a kernel package was
> >>    used. (Closes: #229237)
> > 
> > This does not fix #229237: lilo and grub do not update the atime of
> > the kernel you boot, so this is useless. I strongly suggest we
> > revert it.
> 
> Ah, right.  Good catch.  My testing was flawed.  I noticed that the
> kernel package did get some usage time, and did not check if it
> matched the boot time.  But who said the vote for kernel packages
> should match the boot time?  As far as I know, the point of skipping
> some files in the packages is to avoid files which are accessed by
> cron jobs, and none of the files in /boot/ are accessed by any cron
> jobs I know about, so why should we not look at the files in /boot/?

So when the atime get actually updated ?
If we do not know how to interpret the data, the data is useless. 
This is a loss of privacy without any benefit.

> > On the other hand, it seems that kernel modules atime are updated so
> > maybe this is a solution, though depmod will reset them, which would
> > render the atime useless. This needs to be investigated, though.
> 
> Hm, interesting point.  Not sure if there are any other useful ways to
> get votes from kernel packages.
> 
> > As long as we keep SMTP as a back up, I have no objection, but I
> > would like to see more data before claiming that HTTP submissions is
> > generally reliable.
> 
> What kind of data do you want to see?  The majority of popcon
> submitters are using HTTP at the moment.

What is the proportion of HTTP and SMTP report received ?
I will give you my estimate tomorrow. What is yours ?

> >   * Change package description and question text to make it less
> >     debian specific and avoid indication that only Debian packages are
> >     reported (also non-Debian packages are reported).  Based on patch
> >     from Ubuntu and input from Christian Perrier.
> > 
> > I disagree. popcon report Debian packages and only Debian packages,
> > popcon does not report other types of software packages.  The fact
> > that it reportl Debian packages made by ther distributors does not
> > change that.
> 
> Do you mean '.deb' packages when you write 'Debian packages'?  I read

No. '.deb' files are files embedding a Debian package.

> 'Debian packages' as packages from the Debian archive, and wanted to
> make it clear that also non-debian packages (aka the packages listed
> in the unknown section on <URL:http://popcon.debian.org/>) are
> reported.

The usual usage is that a Debian package is something registered with
the Debian package management system. This is exacly what popcon report.
Packages registered by others systems are not reported.

> > Also popularity-contest, as shipped report to the Debian developers
> > only, not to the distribution developers if you are using another
> > distribution. Claiming otherwise is not accurate.
> 
> Well, it isn't incorrect either, and it do make the description

This _is_ incorrect: if you install popularity-contest from Debian on
Debian Edu, it will report to Debian not to Skolelinux.

> accurate for example for Debian Edu, with the
> <URL:http://popcon.skolelinux.org/popcon/> collector.  I see the
> advantage of reducing the amount of patching required to make the
> description accurate for derived distributions of Debian, and thus
> decided to accept this change.

They have to patch popcon to make it report to another server anyway,
so they should fix the documentation to suit.

> >   * Update standards-version from 3.6.2 to 3.7.2.  No changes needed.
> >   * Add Build-Depends with the same dependencies as
> >     Build-Depends-Indep, to keep lintian happy.
> > 
> > Since this is clearly a bug in lintian, I don't see the point of that.
> 
> Well, it reduces the noise, and do no harm as far as I can see, so I
> changed it to get rid of the message.

It harms non-buggy programs that use Build-Depends/Build-Depends-Indep
correctly.

> Should we try to get a new popcon release out this weekend, or is
> there some changes urgently needed to be implemented first?  I want to
> give the most active translators a day or two, but expect lots of
> translation to show up in the next week, and believe we should
> schedule a new release with updated translations just before base
> freezes.

We should wait a week more. I have still ton of change to review.

I am still disappointed you never ask my opinion _before_ commiting a
change.

Cheers,
Bill.