Awesome! Nice Job.
I would say make a patch for your changes, move to the trunk of
melk.silo.potato and check that in there. Scott tells me your commit
access should be all set. There are no differences between the tag
you're on and the trunk so it should be an easy merge. The rest of
the trunk is not doing anything significantly different with the
backend, so it should be fine to continue to work with the 0.2.1
branch of the frontend until or if you need to modify other
packages. If you feel like living on the edge though, it should be
fine to move onto the trunk entirely.
I can give you a hand switching out that package, but basically you
want to activate your melkjug virtualenv, check out the trunk of
melk.silo.potato in the src/ folder of your installation and run
'python setup.py develop'
Both of your suggestions sound fruitful, I'd say priority would be to
get it to be at least a little smarter about executing the current
interface. Essentially, tune it up a bit if you see some obvious
things. If we could beat the mysql backend for a handful of users on
a single server, that would be a big immediate win. If it doesn't
look like there is much that that can be done in short order (next
week?) or it would require extensive changes to how it's accessed,
write up a description of what's wrong about how we're doing it and
outline the work that would need to be done as you see it instead.
Spend some time on the clustering too, if it looks like it's easy to
get up and running, that's fine, but don't focus on it to the
exclusion of the other stuff. In either case write a little something
up about how hard or easy it is, how well it worked etc. I'm not sure
you need vms to cluster? It seems like you may only need to run the
server on a couple of different ports, at least to figure out how it
works.
I would like to get you working on the view-drives-filters idea after
you have taken a quick pass at the current thing. I would say keep
that in mind when thinking about improving / restructuring the current
one. Getting the python view server going and how to get it's
environment right would also be something to toy around with in the
mean time if you get bored.
- Luke
On Jun 12, 2008, at 6:23 PM, Randall Leeds wrote:
After some tweaking of the way Melkjug updates CouchDB views, I'm
pleased to say that the test suite for the CouchDB storage silo
passes with flying colors on release 0.2.1.
Question then is:
1) Should I start trying to implement smarter use of CouchDB?
a) try to experiment with clustering (I've been wanting a reason
to run some vms lately...)
b) try to experiment with better views
2) Should I move over to the svn trunk and start working from there
on (1-a) and (1-b)?
-Randall
!DSPAM:4013,4851a3a2288132085621377!
