On 09/23/2011 11:23 AM, David Fetter wrote:
>> I'm having a lot of trouble figuring out what to do about this.
>> Should I just do a minor tweak to that section with a reference or
>> two to PostgreSQL doc or wiki pages, or does it merit something more
>> fundamental?
>>
> As in a pointer to a new wikipedia page on SSI?
>
The page Kevin has issues with already points to the section of a
Wikipedia page that addresses SSI:
http://en.wikipedia.org/wiki/Snapshot_isolation#Making_Snapshot_Isolation_Serializable
I think the right order of operations here is to make sure that's
absolutely correct; then return to
http://en.wikipedia.org/wiki/Isolation_%28database_systems%29#SERIALIZABLE
and just rewrite it to note there is a concrete solution available.
Maybe something like this:
When using non-lock based concurrency control, via an approach such as
[[http://en.wikipedia.org/wiki/Snapshot_isolation#Making_Snapshot_Isolation_Serializable|Serializable
snapshot isolation]], no locks are acquired. If the system detects a
write collision among several concurrent transactions, only one of them
is allowed to commit. See snapshot isolation for more details on this topic.
I don't know what other issues Kevin is alluding to, besides this
reading as if there are no good solutions around--which the above text
should help with.
--
Greg Smith 2ndQuadrant US greg@2ndQuadrant.com Baltimore, MD
PostgreSQL Training, Services, and 24x7 Support www.2ndQuadrant.us
