Re: Minutes of the Board Meeting of March 11th, 2014

[Richard Hughes <hughsient gmail com>]

On 25 March 2014 22:41, Richard Stallman <rms gnu org> wrote:
> We might _store_ just a hash code, but I am pretty sure the messages
> contain the real IP address of the sender.

Sure, just as much as your IP is sent to the NTP server or the
google.com homepage.

> The feature should be useful, but it should operate locally.  Based on
> whoever chooses to tell us what is installed, we could calculate a set
> of recommendation rules.  The package installation facility could
> check those rules to make recommendations, in a purely local way.
> That would avoid asking people to upload their list of installed packages
> in order to get recommendations.

That's exactly what I want to do, getting recommendations from a
remote source would be both inefficient and unnecessary. Your
arguments identifies the root cause of all this effort; without _data_
we can't actually generate these client side rules. I could manually
go though a thousand applications and try to think what *I* might use
if I had this application installed, but that's going to be both
wildly inaccurate given I'm not the kind of person to generate
knitting patterns or listen to MIDI music. It's the data I need.

Ohh, and asking random people to send me a package list is going to
wildly skew the results; normal people don't run random commands in a
terminal. We want data from people using GNOME, rather than
*developing* GNOME. If you can think of a better way to get this data,
I'm all ears.

Richard