01:42:03 <stepcut> HugoDaniel: Well.  we are staring with kdtrees to see if we can make that work.. they seems to work better than the current IxSet. But we can always try other things as well
01:42:36 <stepcut> HugoDaniel: one nice thing about MACID is that you can use whatever datatypes you want.. so maybe we will have several IxSet-like datastructures that are optimized for different types of access patterns
01:42:55 <HugoDaniel> :)
01:42:59 <HugoDaniel> that is great
01:43:04 <stepcut> HugoDaniel: I think it is pretty clear that there is not one implementation that is optimal for all operations
01:43:51 <HugoDaniel> are you planing on replicating data between those data structures ?
01:44:19 <stepcut> that is one of the selling points of MACID. Rather than say you have to make everything into table based relations, or everything into key/value pairs, etc. We instead give you a general purpose persistence mechanisms so that you can use whatever datastructure actually makes sense
01:45:05 <HugoDaniel> hmm hmm
01:45:25 <stepcut> when replicating between servers, you would probably use the same datastructure on all the servers. But, for migration, you can migrate from one variant to another. For IxSet and similar types, we just convert the data to a list and serialize the list
01:45:32 <stepcut> so when you read the list back in, you can convert it to whatever you want
01:46:03 <stepcut> ok, time to eat this wings before they get  cold
01:47:53 <HugoDaniel> :)
01:53:36 <HugoDaniel> http://hpaste.org/44714/databasedatatypes
01:53:50 <HugoDaniel> this is the datatypes for a simple database abstraction i did