[sldev] Cache speed experiment & results...

[darien_caldwell at comcast.net]

This reminds me of something I once pondered. Do you think the UUIDs are random enough, that one could rely on a comparison of only the first 8 digits as a sort of quick and dirty identification? What would the chance be of a false match in such a case?


 -------------- Original message ----------------------
From: Alex Dailey <alex at lindenlab.com>
> A humorous side note to this. The distribution of uuid's is so random 
> and there is so much 'space' between values, that Cory et al. joked 
> about putting phony assets in the early range, ...0000000, 0000001 that 
> rezz'ed signs saying things like, "Give up now!" and "Abandon your quest 
> or despair!" for people who tried to guess asset_id's starting from the 
> beginning of the uuid range.
> 
> Kent Quirk (Q Linden) wrote:
> >
> > On Jun 4, 2008, at 12:54 PM, Celierra Darling wrote:
> >
> >> On Wed, Jun 4, 2008 at 12:37 PM, Dale Mahalko <dmahalko at gmail.com> 
> >> wrote:
> >>> My only concern is that the GUID distribution is may be just merely
> >>> pseudo-random.  ...
> >>
> >> Why not hash the (entire) GUID and use that instead?
> >
> > The point of a GUID is that it's already random. Or more correctly, 
> > "merely" pseudo-random. Not that most computers have a handy source of 
> > true randomness anyway.
> >
> > Our GUID generation code is found in lluuid.h/.cpp. There are comments 
> > in there that indicate that we've thought about distribution and the 
> > quality of our randomness before.
> >
> > Which makes sense, given that we have to store *all* of those assets 
> > ourselves. If they had a tendency to sort themselves into buckets, 
> > we'd have a lot more trouble managing them than any external caching 
> > scheme will. I don't think this is something to bother with.
> >
> >     Q
> > _______________________________________________
> > Click here to unsubscribe or manage your list subscription:
> > /index.html
> 
> _______________________________________________
> Click here to unsubscribe or manage your list subscription:
> /index.html