Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Simon Riggs <simon@2ndQuadrant.com> writes:
>> On Fri, 2011-01-21 at 11:19 +0200, Heikki Linnakangas wrote:
>>> It's not the order in which the xid was assigned that matters,
>>> but the order the transactions started and got their snapshots.
>>> The xids might be assigned a lot later, after the transactions
>>> have already read data.
>
>> So if a read-write transaction assigns an xid before it takes a
>> snapshot then we'll be OK? That seems much easier to arrange than
>> passing chunks of snapshot data backwards and forwards.
We're not talking about passing the backwards. I'm suggesting that
we probably don't even need to pass them forward, but that
suggestion has been pretty handwavy so far. I guess I should fill
it out, because everyone's been ignoring it so far.
> No, that idea is DOA from a performance standpoint. We sweated
> blood to avoid having to assign XIDs to read-only transactions,
> and we're not going back. If SSI requires that, SSI is not
> getting committed.
SSI doesn't require that. The suggestion that it would in *any* way
help with the interaction with hot standby is off-base.
-Kevin