I decided to work against this urge now and work incrementally for where we are. I have already written the DTDs and the validation scripts in the gump CVS module and forrest is the only one not validating.
The validation is not really useful since the markup is so freeform, but it's better than nothing already.
So, next problem is the noisy data visualization.
There are many ways to implement this but the easiest and more XP one is to just reuse what Adam already coded. So, I installed WebWare on brutus (webware is like servlet API for python) and I'm going to work on having "dynagump" simply using the same classes that Adam wrote to read the workspace into memory and then query the history database to know what happened and when.
So, this is what happens
1) cron runs gump, which read the workspace into memory and uses it to launch the builders, update the historical database and send the nagging reports
2) the webapp uses the same code to read the workspace in memory, uses it to create a visualization of the dependency graph and hits the build history database to show results (also allows to do more general things like time graphs and dynamic FOG statistics).
This is not the best possible architecture, but it's the simplest to implement right now.
And this is what I decided to work on. No java, no cocoon, no rewrite, lazy ass reuse of what's already there.
Let's see if this approach works better ;-)
-- Stefano.
smime.p7s
Description: S/MIME Cryptographic Signature
