From: Jeff Bone (jbone@jump.net)
Date: Mon May 07 2001 - 14:00:11 PDT
Eugene.Leitl@lrz.uni-muenchen.de wrote:
> I'm not sure caches have any point at all where you have many spatially
> distributed agents (fancy that, since we can't make purely bosonic computers)
> simultaneously mangling data, which must remain consistent.
Again, the question is: is strict consistency for some arbitrary definition of "simultaneous" a
requirement of the app, or is eventual consistency (or even no consistency) within some window
acceptable? My contention is that the latter is acceptable for most applications.
> Anyway, the assumption of absolute consistency is another human artifact.
Agreed.
> Anyway, the destillation of this crappy essence is: the first rule of caching: don't.
Well, see, I disagree with this and I believe your argument is now internally inconsistent. You're using
biological examples, so cases in point: cross-regulation of protein production, genetics, squirrel nut
caching strategies, endocrine control, maintenance of homeostasis, population self-thinning in trees,
migratory bird formations, ant foraging behavior, group "flight" behaviors in the presence of wildfire,
(apparently) storage of memories and abilities within animal brains, and countless other biological
examples all rely on local, incomplete, and potentially-inconsistent information, i.e. caching with weak
consistency. Caching happens *everywhere* in nature; biology is *never* single-point-of-failure in any
global sense, and it *never* requires global coordination and consistency. The world is messy; even
physics *cannot* require strict global consistency of complete information (Heisenberg, Bell, Einstein,
et. al.) It's a peculiarly-human psychological tendency to think that consistency is at all important in
systems.
So we agree on the importance of biological models and metaphors, but I'm baffled by how you can
therefore conclude that caching is a "don't." IMO, working from those biological metaphors, it's a
"must" for decentralized, loosely-coupled systems of any significant complexity.
> Do we still agree? ;)
Well, we did until you made an inconsistent argument...
;-)
jb
This archive was generated by hypermail 2b29 : Mon May 07 2001 - 14:49:58 PDT