gump-general mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Leo Simons <leosim...@apache.org>
Subject Re: How to detect "Friends of Gump"
Date Tue, 08 Apr 2003 07:59:27 GMT
Stefano Mazzocchi wrote:
> I've been working more and more with gump and I've started to grow an
> itch to differentiate those community who *care* about gump and those
> who don't.

it is not black-and-white. Not even grey. It's more like a box of 
chocolates :D

> - create a (simple yet meaningful) statistical model that will analize
> the trend of 'green/red/yellow' status of each project to indicate its
> 'gump friendlyness'.

I don't think it is reliably possible to define such a model. There are 
many many factors to throw into the equation (like internal project 
complexity, the amount of interaction with outside projects, etc etc, 
but also the lack of support for parts of a project build process, 
whether or not a projects runs its unit tests, whether it becomes red on 
a test failure, ....).

It would still be nice to have the existing statistics graphed. 
Interesting and measurable quantities would be

size
----
1 - the number of projects in gump (vs time)
2 - the size of the codebase of projects in gump (vs time)
3 - the time it takes to do a gump run (vs time)
4 - the number of gump failures (vs time)
5 - the number of gump prereq failures (vs time)
6 - the number of successfully built gump packages (1-4-5)
7 - gump uptime
...

change
------
besides the change of any of the quantities above...

6 - the average time to fix a failure (vs time)
7 - the average time to fix a prereq failure (vs time)

accelleration
-------------
as always the change of change...


of course one could do the same for a single project, a group of 
projects, a cvs module, a group of modules, but I think the value of 
those results would diminish rather quickly. As another example, it is 
rather bad for gump if avalon-logkit or avalon-framework fails, so a 
failure there I will usually make an immediate priority. This would lead 
to a low (6). However, failure of excalibur-fotress-examples is a lot 
less important, as the success or failure of that package only affects 
the package itself, and it is unreleased software. As a result, "the 
avalon community" can probably be considered 'friend of gump' when 
looking at avalon-framework (which also happens to be a stable package, 
so it is easier to maintain), but not when looking at 
excalibur-fortress-examples.

Summary: yay for collecting and visualizing stats, sceptical wrt a 
statistical model (but feel free to amaze me!).

cheers!

- Leo



Mime
View raw message