[jira] Commented: (RIVER-84) Outrigger could resolve transactions atomically

["Fred Oliver (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/RIVER-84?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12515876 ] 

Fred Oliver commented on RIVER-84:
----------------------------------

 If bug 5054742 - "Transaction spec does not specify whether or
not participant commit is atomic" is resolved to require atomic commit and
abort then this should become a bug and needs to be fixed. If it get
resolved to not require atomic commit and abort we may not want to change
Outrigger so it does not go beyond what is required by the spec.

Note: 5054742 appears to have been resolved with these comments:
- Explicitly add "weak" language to the default Transaction specification. That is, allow participants to have un-atomic lock release semantics, which could cause isolation to violated by another observer (like the provided scenario).
- Don't preclude specific participants from providing stronger semantics, though. 

> Outrigger could resolve transactions atomically
> -----------------------------------------------
>
>                 Key: RIVER-84
>                 URL: https://issues.apache.org/jira/browse/RIVER-84
>             Project: River
>          Issue Type: Improvement
>          Components: com_sun_jini_outrigger
>    Affects Versions: jtsk_1.0
>            Reporter: Fred Oliver
>            Priority: Trivial
>
> Bugtraq ID [5054755|http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=5054755]
>  Currently if you started with an empty space and you had two clients where :
>    Client 1 writes under transaction T entries E1 and E2 that both
>    match the template S.
>    Client 1 calls commit on T.
>    Before the commit on T returns client 2 calls 
>       space.take(S, null, 0) and gets E1.
>    After the first take returns, but still before
>    the commit on T returns client 2 calls 
>       space.take(S, null, 0).
> Outrigger could return null from the second take. This makes visible that
> Outrigger does not atomically commit transactions. I have done a full
> investigation, but this would be pretty easy to fix in Outrigger.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.