[sldev] Cache speed experiment & results...

[Robin Cornelius]

Kent Quirk (Q Linden) wrote:

> There are a couple more issues here:
> 
> a) UUIDs are constructed from a collection of bytes, some of which are a
> timestamp. Using a subset of them might well lead to collisions.
> b) The probability of a collision if you only use 8 bits is about 1.1%
> if you have 10K items, and about 68% if you have 100K items. That seems
> WAY too high to me. It's the birthday "paradox" -- the probability that
> two people have the same birthday. In python, the formula is:

Yes of cause and thanks soft for the link too. I only considered a
single match which was a straight probability but we are considering a
match in a pool where any member can match any other member so the
actual resultant probability drops substantially in relation to the
square of the pool size.  I did warn you about my math(s) ;-)


-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 252 bytes
Desc: OpenPGP digital signature
Url : http://lists.secondlife.com/pipermail/sldev/attachments/20080604/449d7649/signature.pgp