Re: [boost] [Reminder][Review] The review of the Boost.Guid library

From: Andy (atompkins_at_[hidden])
Date: 2007-05-08 09:53:59

<Oliver.Kowalke_at_[hidden]> wrote in
news:B1EAFF2DAE7658488B631F25813CD91F0137D996_at_[hidden]: 

>> > How useful can seed using MAC address + time be in system with 
>> > multiple cores? Shouldn't it be CoreID + time? Just image 
>> you have 16 
>> > cores and just one MAC address. Is  such an assumption 
>> valid to make?
>  
>> You would need process/thread number + mac address + time
> 
> Hmm - the RFC 4122 doesn't tell anything about threads.

This is not exactly true.  The first step in their basic algorithm
(4.2.1) is to obtain a system-wide global lock.  This lock ensures that
each guid has a different time stamp.  Often an implementation strays
from the standard by adding in a process/thread/core number to avoid a
global lock.  It also talks about an optimization (in 4.2.1.4) that
allocates a block of time stamps at a time and a per-process generator
can allocate from that block until it is exhausted. 

> Version 1 UUIDs are generated according to the following algorithm:
> 
>    o  Determine the values for the UTC-based timestamp and clock
>       sequence to be used in the UUID, as described in Section 4.2.1.
> 

< snip >

> 
> Oliver

Andy.