Simon Gray wrote:
> Paul J Stevens wrote:
>> But there is another reason in the imap requirements: UID values must be
>> strictly ascending within a mailbox.
>>   
> Rather than making the UID just an autoincrement, why not make it a
> function with time? Therefore as long as the servers are time sync'd
> (easy with ntp) you should be able to increment the number based on the
> time, perhaps there is some way to concatenate the date? E.g.
> 
> To have
> Server A odd auto_inc
> Server B even auto_inc
> 
> new uid = sql concat(unix_timestamp(), UID)
> 
> Therefore they should always increment and be unique between servers?
> 
> Worth a thought or completely silly?

Worth a thought *and* not-quite-silly but it still wont do. In fact I did an
implementation of this once. There will still be ample opportunity for one
server to store a UID that is below the MAX(uid). *Any* chance of such an
occurence is a fatal blow to your algorithm.

My more fundamental conclusion was that the whole master-master dream is a myth:
it doesn't solve anything. It doesn't scale at all.

Does it offer higher throughput?
Does it offer lower latencies?
Does it offer higher reliablity?
Does it offer higher availability?

think about it.

-- 
  ________________________________________________________________
  Paul Stevens                                      paul at nfg.nl
  NET FACILITIES GROUP                     GPG/PGP: 1024D/11F8CD31
  The Netherlands________________________________http://www.nfg.nl
_______________________________________________
DBmail mailing list
[email protected]
https://mailman.fastxs.nl/mailman/listinfo/dbmail

Reply via email to