[jira] Commented: (JAMES-913) Incorrect MX record orderering

["Norman Maurer (JIRA)" <se...@james.apache.org>]

    [ https://issues.apache.org/jira/browse/JAMES-913?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12805556#action_12805556 ] 

Norman Maurer commented on JAMES-913:
-------------------------------------

Anyone has some idea howto fix this ?

> Incorrect MX record orderering
> ------------------------------
>
>                 Key: JAMES-913
>                 URL: https://issues.apache.org/jira/browse/JAMES-913
>             Project: JAMES Server
>          Issue Type: Bug
>          Components: DNSServer
>    Affects Versions: 2.3.1
>            Reporter: Amichai Rothman
>            Assignee: Norman Maurer
>            Priority: Minor
>
> The SMTP RFC states that given multiple MX record, they must be ordered such that lower priority records come first, and equal priority records are randomized for load balancing. The JAMES implementation uses Arrays.sort() along with an MXRecordComparator, which compares the priorities, and 'flips a coin' if they are equal.
> This precariously breaks the Comparator contract, which can theoretically cause a sort implementation to loop forever, for example if it is implemented using a trivial bubble sort which stop when no more swaps are performed. In practice, it causes a different bug: the Sun implementation of Arrays.sort() uses insertion sort for N <= 7. So for example, if you have 5 MX records with equal priority, the last element is compared only in the last iteration of the loop, with a 1/2 probability of remaining last. On the other hand, it has a 1/16 probability of being swapped 4 times to become first on the list.
> So while there is randomness involved, equal-priority records will tend to stay near their original (relative) position, and not be shuffled with the uniform probability which the RFC implied for load balancing...

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


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