journal spec?

[Ivan Krstić]

Joshua N Pritikin wrote:
> I don't know what has been discussed so I'll just blurt out my idea. 
> Based on the description of the Journal in the HIG, it seemed like GIT 
> would be a close fit for the database requirements. 

That was the first thing I thought about. No dice; a generic VCS store
won't cut it for us. git relies on an immutable history and is (almost)
exclusively content-addressable, neither of which is acceptable for us.

I've been spending half my time on this for almost two months, and it's
turned out to be a very difficult problem, with one full implementation
already thrown away. What we want is most akin to an RDF store, but
those are nowhere near performant enough for us to use (certainly not
the free and open ones, anyway).

Much of the final requirements were only hashed out in meetings over the
last 2 weeks, and I don't have this written up anywhere except in my
design notebook (the paper kind). I'm racing now to release a working
implementation this week from which we could iterate further, and will
try to release it with full docs and a spec that incorporates the latest
thinking.

-- 
Ivan Krstić <krstic at solarsail.hcs.harvard.edu> | GPG: 0x147C722D