[cap-talk] Is "Authority" Subjective?

[David Hopwood]

Pierre THIERRY wrote:
> Scribit Jonathan S. Shapiro dies 22/06/2007 hora 13:35:
>> Not generally. Yes, there is concurrency in the real system, but the
>> *model* generally assumes a layer at which execution proceeds by some
>> serializable sequence of atomic operations.
>>
>> The serialization need not be unique, but at least one such
>> serialization has to exist.
> 
> But when the serialization is not unique, you cannot reason on it.

That's not true. The existence of a serialization is often sufficient.
It's the basis for consistency in transactional databases, for instance.


(All commercial database systems support serializable isolation, although
they may also support weaker isolation levels in the name of performance.
Unfortunately the ANSI standard describing isolation levels for SQL has
more influence than it deserves: see
<http://citeseer.ist.psu.edu/berenson95critique.html> for some
criticisms of it. I haven't seen more than handwaving about the actual
effect of isolation level on database performance.)

-- 
David Hopwood <david.hopwood at industrial-designers.co.uk>