commons-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Tim Reilly" <tim.rei...@consultant.com>
Subject RE: [PROPOSAL] [id] Prefix, UUID and Host StringIdentifierGenerator
Date Fri, 13 Feb 2004 00:13:12 GMT
Sorry I've been MIA for the last week or two. I'm making some tentative
changes to the UUIDClock contract basically in order to accommodate systems
with multiple node identifiers I've realized that the Clock getUUIDTime
contract likely needs to be: returns a unique uuid timestamp within
consecutive calls per Node identifier (e.g. mac-address)

- The reason I qualify this with consecutive calls; is due to the fact that
the system time could move backwards due to a synchronization with an NTP
service or someone or something adjusting the systems clock backwards. (In
which case a new clock sequence needs to be created.)

- The reason to make this per Node identifier is because uniqueness of the
timestamp is critical for a particular node identifier + clock sequence
combination. If more than one node is present for the JVM then more uuids
can be generated for a given system time interval than is possible by using
only a single node identifier.

Alban - I wanted to give you a heads up regarding the possible change. If
the changes are not compatible with your work then perhaps some sort of
timestamp identifier implementation/generator not tied to the uuid generator
could be possible?

(Also I wanted let Phil, Gary, et al know I've not vanished ;-)

I'll hopefully have changes to post to bugzilla within the next few days.
Plus some documentation regarding the changes and see what everyone thinks.

-TR


---------------------------------------------------------------------
To unsubscribe, e-mail: commons-dev-unsubscribe@jakarta.apache.org
For additional commands, e-mail: commons-dev-help@jakarta.apache.org


Mime
View raw message