On Tue, Feb 27, 2018 at 5:07 PM, Peter Geoghegan <pg@bowt.ie> wrote:
> I now feel like Simon's suggestion of throwing an error in corner
> cases isn't so bad. It still seems like we could do better, but the
> more I think about it, the less that seems like a cop-out. My reasons
> are:
I still think we really ought to try not to add a new class of error.
> * We can all agree that *not* raising an error in the specific way
> Simon proposes is possible, somehow or other. We also all agree that
> avoiding the broader category of RC errors can only be taken so far
> (e.g. in any event duplicate violations errors are entirely possible,
> in RC mode, when a MERGE inserts a row). So this is a question of what
> exact middle ground to take. Neither of the two extremes (throwing an
> error on the first sign of a RC conflict, and magically preventing
> concurrency anomalies) are actually on the table.
Just because there's no certainty about which behavior is best doesn't
mean that none of them are better than throwing an error.
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company