On Mon, 2013-06-03 at 17:08 +0200, Mathieu Stumpf wrote:
> Le 2013-06-03 16:38, Bastien Nocera a écrit :
> > On Mon, 2013-06-03 at 15:32 +0200, Mathieu Stumpf wrote:
> >> Le 2013-06-03 07:48, Luc Pionchon a écrit :
> >> > Having “real life” datasets help to see how the software scales in
> >> > real environment.
> >> >
> >> > In the case of an Artist list, as an example, you may have 50 
> >> artists
> >> > in your own music library. *Plus*, if you have 10 compilations 
> >> (movie
> >> > soundtrack? dance compilation?) with each 30 individual artists , 
> >> and
> >> > tadam… +300, your artist list is flooded with artists (whom you 
> >> never
> >> > heard) with a single track. Your artist list becomes useless,
> >> > although
> >> > it looked so nice and clean and easy on the mockups.
> >>
> >> There are plenty of good free/libre songs out there, you know?[1]
> >
> > Except that you don't want to download 100 GB of songs to test a
> > software. Blank songs would compress to almost nothing, and more or 
> > less
> > to the size of the metadata once tarball'ed up. It should be possible 
> > to
> > offer tens of thousands of songs with a download size in megs, not 
> > gigs.
> >
> > The construction of good test cases with the metdata is more 
> > important
> > than the contents of the files.
> 
> Hopefuly, you mean that only in the very specific case of a software 
> design process. ;)

I have well-tagged awful music ;)


_______________________________________________
usability mailing list
[email protected]
https://mail.gnome.org/mailman/listinfo/usability

Reply via email to