[sldev] Cache speed experiment & results...

[Argent]

On 6/4/08, Dale Mahalko <dmahalko at gmail.com> wrote:
> My only concern is that the GUID distribution is may be just merely
> pseudo-random.  When this system was dreamed up at the start of SL,
> did anyone actually do studies of the randomness of the distributions?

I believe the SL UUIDs are actually generated randomly, often on the
user's computer(!), with a variety of seeds.

But if that turns out to be an issue... don't forget that this is just
a cache, there's no issue with *changing it* to take other bits from
the UUID for the first couple of levels. You could even use a simple
CRC or other cheap polynomial on the UUID to generate a 16-bit value
for the first two levels.

If the UUID is still used for the rest of the file name, you don't
even need to version the cache, all that will happen is you'll get
cache misses when the algorithm changes and the old files will
eventually age out of the cache.