[fab] Re: splash page + tracking image example

[Stephen John Smoogen]

On 10/5/06, Max Spevack <mspevack at redhat.com> wrote:
> On Thu, 5 Oct 2006, Josh Boyer wrote:
>
> > No, it's not.  DHCP, NAT, proxies, etc. all make "unique" IP addresses a
> > horrible metric.  In some cases, you get a single IP address for a
> > number of different machines.  In others, you get two IP addresses for
> > the same machine.
>
> That's true, it will not be perfect by any means.
>
> > Yes, counting IP addresses will give you a metric.  I just don't think
> > that metric is good for much because you still don't know how many
> > machines really have Fedora installed.  Is it better than nothing?  I
> > don't know.
>
> It gets you a pile of data, which can be analyzed and from which value can
> be extracted.

As long as you put in error bar estimates on what you are analyzing.
When I was collecting numbers for RH Marketing.. the big problem I had
was educating people that the numbers I gave were not absolutes and
had errors involved on it. Not knowing that there are estimates in the
numbers were causing problems where website growth was occuring in
certain countries and such.

The issue with any of the tracking schemes is working out where you
are not violating a persons privacy. A person who is doing updates on
the internet is sort of making an agreement that they can be counted..
but it may not be one that 99% of the people doing it realize. [The
remaining 1% either dont care or use Tor or something to mask
themselves.]

To make it work you need to make an agreement with people. A simple
choice would be to put in a firstboot page in FC7. State basically
what Greg did why we would like to track certain information and how
it is important for future of Fedora. Then if they agree to it, then
send out a webcookie for the machine that is made from a SHA1 hash of
some amount of random data. Put that data also into a yum plugin that
sends the data as a cookie that is in the yum sned to the master
server.

This allows you to collect a certain amount of data and clear out some
of the issues about dhcp, etc. People who do not want to participate
don't put the cookie in their send, those that do can give some
information that isnt too privacy threatening. [ The too is meant to
cover people who believe that there is absolute right to all privacy.]

>
> It's certainly better than having *no data*.
>
> Bottom line -- in th FC6 timeframe, any metric gathering will be
> imperfect, and feel like a hack.
>
> So we pick the hack that won't also anger people, and we do it.
>
> --
> Max Spevack
> + http://fedoraproject.org/wiki/MaxSpevack
> + gpg key -- http://spevack.org/max.asc
> + fingerprint -- CD52 5E72 369B B00D 9E9A 773E 2FDB CB46 5A17 CF21
>
> _______________________________________________
> fedora-advisory-board mailing list
> fedora-advisory-board at redhat.com
> http://www.redhat.com/mailman/listinfo/fedora-advisory-board
>


-- 
Stephen J Smoogen. -- CSIRT/Linux System Administrator
How far that little candle throws his beams! So shines a good deed
in a naughty world. = Shakespeare. "The Merchant of Venice"