[Dev] uuid could/should be simplified

Roger Eaton RogerEaton at brandx.net
Tue May 18 09:08:10 PDT 2004


Thanks for answering in such detail, Andi.  It sounds like in the end 
you might go with
all random uuid's, except that it is not as easy to do as it seems.

I don't see the need for "cryptographically secure" so long as the 
process covers the
space evenly without generating duplicates more often than expected by 
chance.

I am planning to just use the built in Python random function, seeded 
with something
like time plus hash of the contents of the first item to be added to the 
InterMix repository
on this run plus hash of contents of previous item added.   The write up 
of the random
function in the Python library reference says that it is very good at 
evenly covering the
space and is fast.  Won't this work? 

 > So, what's the bug here, am I missing something here that needs to be
 > addressed before Year 5238 ?

No bug -- my mistake.  Now I see -- you dont overlay the high order bits 
of the
date because the date doesn't get that big until 5000 and something.  I 
was thrown off
by the code that zeros the high 4 bits -- uuid[6] = 0x10 | (uuid[6] & 
0x0f); but
those hi bits are always zero to begin with.

-- Roger Eaton
rogereaton at brandx.net
http://blog.voiceofhumanity.net




More information about the Dev mailing list