Cyril SCETBON <cscetbon.ext at orange-ftgroup.com> writes:
> Jan Wieck wrote:
>> On 9/10/2007 12:49 PM, Cyril SCETBON wrote:
>>>
>>> Christopher Browne wrote:
>> Turning Slony into a statement based replication system would make
>> it far worse.
> You may be right. It was just a suggestion to do something to replace
> the long SQL string by a shorter one to gain space, and so performance
> when getting blocks from disks with more rows :-)
>> Queries depend on the exact transaction commit order with respect to
>> the exact rows that are visible to them. That in addition to the
>> non-deterministic behavior of sequential scans and a myriad of
>> functions ... how do you expect some query like
>>
>>     update mytable set foo = nextval('someseq') where foo is null;
>>
>> to lead to exactly the same result on two different databases? Not
>> even the order in which the rows are processed is defined.

> If MySQL support it , why slony would not ? 
> (http://dev.mysql.com/doc/refman/5.1/en/replication-features-autoincid.html),
> but you right for deterministic functions it's hard if value is added
> in the statement. But it's not what I've suggested.

Just because they choose to jump off a cliff, that doesn't make it
make sense for us to do so.

Your suggestion does indeed amount to asking for Slony-I to behave in
the same sort of nondeterministic mode that the MySQL replication
system sometimes suffers from.

>> Compared to Postgres, MySQL is a fairly limited system feature
>> wise. Yet with their limited functionality the MySQL user manual
>> already states that several features are not supported by their
>> statement based replication. What do you think the corresponding
>> documentation for Slony would look like?

> Sorry, but you can compare it to the log shipping feature of slon
> with less lag time.

You can only do that if you also commit to wishful thinking:

 "If I choose not to care about data integrity, we can make the system
  somewhat more efficient."

That is decidedly NOT one of the design considerations for Slony-I.
One of the express goals is to provide confidence of data integrity.

If we drop that goal, we will lose the existing user base, so that is
a totally unacceptable approach.
-- 
output = ("cbbrowne" "@" "acm.org")
http://www3.sympatico.ca/cbbrowne/wp.html
The statistics on  sanity are that one out of  every four Americans is
suffering from some  form of mental illness. Think  of your three best
friends. If they're okay, then it's you. -- Rita Mae Brown