[
https://issues.apache.org/jira/browse/CASSANDRA-2067?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12988125#action_12988125
]
Jonathan Ellis commented on CASSANDRA-2067:
-------------------------------------------
- message digest instantiation is expensive and synchronized, so we made a
threadlocal for that in FBUtilities. Should probably use that.
- I don't understand the purpose of clockOffsetTicker. it's only used during
the constructor, but constructor is only used once so clockOffsetTicker has no
effect.
- i'd be happier if we used an AtomicLong instead of synchronization to protect
lastNanos. (look at java Random class for an example of using AtomicLong for a
similar purpose)
> refactor o.a.c.utils.UUIDGen to allow creating type 1 UUIDs for a given time
> ----------------------------------------------------------------------------
>
> Key: CASSANDRA-2067
> URL: https://issues.apache.org/jira/browse/CASSANDRA-2067
> Project: Cassandra
> Issue Type: Bug
> Components: Core
> Reporter: Eric Evans
> Assignee: Eric Evans
> Fix For: 0.8
>
> Attachments:
> v1-0001-CASSANDRA-2067-o.a.c.utils.UUIDGen-adapted-from-flewto.txt,
> v1-0002-eliminate-usage-of-JUG-for-UUIDs.txt,
> v1-0003-remove-JUG-jar-and-references.txt
>
>
> CASSANDRA-2027 creates the need to generate type 1 UUIDs using arbitrary
> date/times. IMO, this would be a good opportunity to replace
> o.a.c.utils.UUIDGen with the class that Gary Dusbabek wrote for Flewton
> (https://github.com/flewton/flewton/blob/master/src/com/rackspace/flewton/util/UUIDGen.java),
> which is better/more comprehensive. We can even eliminate the dependency on
> JUG.
> Patches to follow.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
