Hello Vadim,
Tuesday, July 06, 1999 you wrote:
V> There is no problems with UPDATE: updated tuple points to newer
V> version, so we can avoid update of referencing tuples here.
V> VACUUM would have to update referencing tuples (via normal
V> heap_replace, nothing special) while removing old versions.
V> This may cause deadlocks but we could give vacuum higher priority
V> and abort others.
V> So, vacuum is the worst case, as pointed by Tom.
V> No problems with MVCC and other things.
So. The main drawback is higher priority for VACUUM. Not
too large, eh?
When you will decide - to implement or not to implement,
I urge you to think again about the relief on optimizer,
which I stressed many times. No one rebutted yet that adding
brains to optimizer so that it can use appropriate join method
will require major rewrite. With links you get the best join
method as side effect - virtually for free. These joins
will never be too slow for an unknown reason. Think carefully.
I hope you will make wise decision.
Best regards, Leon