Hi All,

I've just been looking at some of the download statistics for ArgoUML. 
I'll write a few things down here, for the benefit of those who don't
have access to the google analytics account.

The windows installer version of ArgoUML is the most popular download,
and has been downloaded between 3000 and 4000 times per week over the
last 6 months, this is gradually increasing.

The download statistics also tell us how many of each alpha/beta version
have been downloaded.

0.28.ALPHA_1 9
0.28.ALPHA_2 20
0.28.ALPHA_3 21
0.28.ALPHA_4 21
0.28.ALPHA_5 30
0.28.ALPHA_6 25
0.28.ALPHA_7 21
0.28.BETA_1 59
0.28.BETA_2 667
0.28.BETA_3 252
0.28.BETA_4 28

Of these, again, the windows version is the most popular, but there also
quite a lot of zip/jnlp downloads of these releases.

I notice that there are two locations where the files are downloaded
from for each release e.g.:

argouml.tigris.org/DOWNLOAD/ArgoUML-0.28.BETA_2...
argouml-downloads.tigris.org/DOWNLOAD/ArgoUML-0.28.BETA_2...

I am not sure why there are two locations... perhaps they point to the
same place.

Observations:
1) The number of downloads rapidly increased when we went to the beta
phase.  I expect that this was due to us putting the link on the front
page.  If this was the reason, then it works, so we should keep doing
this!

2) There weren't many downloads of each alpha release.


I have a suggestion:  If we make the assumption that for every download,
someone actually installs and runs the program, then the amount of
testing could be considered roughly proportional to the number of
downloads.  Perhaps then, we should define our beta period(s), not by a
length of time, but by a number of downloads.  This would be a more
reliable way of ensuring quality.  For example, we could only go from
BETA_n to final once we'd had 1000 downloads of BETA_n.

Anyway, it's something to consider for 0.30.

Regards,

Dave

Reply via email to