Whenever I sit down and thinking properly about this space I always end up thinking about business intelligence (BI). The company I work for has an overwhelming requirement for large scale project measurement and reporting and the problems that need to be addressed are the same one that BI attempts to provide for. Way back when there was a @dev discussion about using a backend DB to store measures accrued through maven plugins and an API to slice and dice that data and represent it in pretty maven site style reports (Vincent me thinks, pages should still be codehaus confluence).
However this approach adds a new independent repository and a whole bunch of complexity to the very clean maven world (I love my 'and here I delete everything and watch maven reconstitute the entire environment as part of the build' bit of my demos). I love the wagon and I would like to use the maven repository/artifact system wherever possible but for this requirement I feel it is just not going to fly. Note: I should caveat that a bit, simple trend and snapshot data could be stored in the repository, it could even be in the form of a database style data source such as hibernate data files and that could be processed but my point is that if I have 500 projects over 4 years thats a lot of data. It will need to be centralized and require OLAP like processing for its real value to be leveraged. Also, when you start thinking about the sorts of data you'll mine and present you very quickly come up with the need for some good visualization techniques such as dashboards and drill-down supporting aggregation reports. This kind of reporting UI requires something a bit more sophisticated than flat HTML (for instance the JIRA changes report is there as simple eye candy, you immediately click through to the wonders of JIRA when trying to analyze or interpret the data). Anyway, again we're back into BI space (or at least the persist, analyze, report space). However, in ALM the demos I've been giving recently I actually pull up xradar's demo pages and use that to illustrate how a this information *will* be pulled together into an overall dashboard and reporting system that will enable project oversight and health checks to be performed, I even go as far as to say we will have a maven dashboard, and here I am thinking directly of the xradar approach, just not using XSLT and vector graphics. What we want, and I say we as in my company, is something much more like JIRA in terms of its visual and functional richness but sat in the maven site world. But no I havent sussed how to get that in a pure maven solution yet. It's easy enough to use some JFreeChart stuff and render some maven site friendly pages (hmm, is it in the doxia world?) but the work of storing, processing and returning the data is, for us, for now, going to be handled by a hand cranked metrics database (90% done) and an open-source OLAP/BI solution (http://www.pentaho.org/). We are then going to implement harvesters that will take data from the various data sources, of which maven is obviously one (checkstyle, PMD, surefire, clover/cobertura, ...), along with others such as our build service (quickbuild/anthill/continuum), JIRA, confluence, DOORS, and other propriety business systems. I am hoping that my work on Pentaho will provide a means to bridge the two worlds. For us the maven site *is* the project portal, its a context aware (module/sub-project/code) reporting and service gateway (i.e. links into fisheye, jira, confluence, forums, mailing lists, SCM, etc). So whatever we do with this BI stuff I will be developing a front end to it that sits in the maven site. Once this is done I would hope to be able to find a way to use our maven dashboard and reports functionality against a simpler, POM configured backend store of some kind based upon the maven repository environment. I said hope didnt I? Glad it's an area everyone's thinking about though... :) John Allen -----Original Message----- From: Arnaud Bailly [mailto:[EMAIL PROTECTED] Sent: 06 September 2006 16:24 To: Maven Developers List Subject: Re: Maven central reporting API Jason van Zyl <[EMAIL PROTECTED]> writes: > > I was just commenting on the use of XML. If I were to design a pure > reporting Data API I would probably not use XML at all. I would > probably use find a format that would provide the fastest data > processing capabilities because I would like real-time reporting. RDF is just W3C buzz word for directed labelled graphs. Objects naturally form directed (labelled) graphs. Hence it seems natural to use the normalized query and representation (N3 and XML) of RDF to represent OO data structures. > > very useful, but I would still like to come up with a Data API that > would provide a canonical format for data that could be consumed by a > reporting system we came up with. Maybe that's too much to do with I totally agree with you. RDF is just a way to store and query data but everything is wrapped into an API for use by plugins and reporters. That's why I think the xradar solution is a pis-aller as we say in french. > limited resources but that's the idea I had. I think in the short > term if we're going to use everything that exists then I think XRadar > is the most complete, well thought out solution that is ready for use > right now. > OK. I am ready to try discussing with xradar team and maybe come up with a midway satisfying solution. Could you please contact them ? regards, -- OQube < software engineering \ génie logiciel > Arnaud Bailly, Dr. \web> http://www.oqube.com --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED] --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
