On Thu, Jul 24, 2003 at 08:17:12PM -0400, Greg Wooledge wrote:
> I wish we *would* go back to the simple LRU cache.  It seemed to work
> a lot better than probabilistic caching does.
I actually kind of like pcaching, however I think a solution that
would make everybody happy (without the more complicated caching
algorithms mentioned in other parts of this thread) is instead of
having the cache LRU and the decision whether to cache probabilistic,
Cache everything, but have cache expiration be probabilistic based on
cache density. Apart from keeping the "I downloaded this so it should
be in my cache" people happy, it would also lead to faster
specialisation, as the decision whether to throw away a piece of data
is done later when more information about the current
(datastore-)specialisation is available.

-- 
Frank v Waveren                                      Fingerprint: 21A7 C7F3
[EMAIL PROTECTED]|stack.nl|chello.nl] ICQ#10074100            1FF3 47FF 545C CB53
Public key: hkp://wwwkeys.pgp.net/[EMAIL PROTECTED]            7BD9 09C0 3AC1 6DF2
_______________________________________________
devl mailing list
[EMAIL PROTECTED]
http://hawk.freenetproject.org:8080/cgi-bin/mailman/listinfo/devl

Reply via email to