Sunday, January 20, 2008

River Disappoints Again

I had a desire to see if I could get Jython or JRuby working with Apache River this weekend for some distributed monitoring and task execution I want to try. Need something quick, reliable and with code mobility and ideally can integrate easily with Java. I did see some new updates on the River site which was initially encouraging, but after digging, looks like little has changed. There were no source distributions which was concerning, but I'd used the older Sun distributions of Jini before, so I wasn't scared about building from source. I proceeded to SVN check-out. Sadly, the listed SVN URL doesn't work - there is nothing there.

Walking up the tree to the /asf/ root, there is no sign of the source or the river project. I'm sure it is around somewhere, but seriously, how much less approachable could this project be? This is no way to attract new users. I didn't think it could get worse from the old Sun site; I was wrong. What a shame.

Looks like it's back to Rinda then. I've avoided Rinda in the past for important data given that there is no reliable storage of tuples. At this point, probably easier to bolt that on to Rinda than to pull anything meaningful out of River. Alternatively, maybe a better approach would be to make smarter Rinda clients that can survive a failure of the master ring server with some local persistence and eventual consistency approach.

1 comment:

fuzzy said...

Too bad.

Could always do the inverse - Rinda on JRuby if you had such a need.

Looks like there is an issue though:

Further evidence that the Tuple Space is cursed ? (inside joke) ;)