paulruizendaal wrote:

>
>
>
>
>
>
>
>
>> Details at http://ewh.ieee.org/r1/boston/computer/starkeytalk.html
>>
>
> Unfortunately its too far to attend.
>

Pity. We'd be happy to put you up. And there's a tiny bit of Monty's
designer vodka that you could have.

> From the details page you link to, I get the impression that you will be making three points during the lecture:
>
> [1] Under MVCC/snapshot isolation each connection has its own database state, as it can 'see' its own uncommited changes
> [2] In a distributed database non-conflicting transactions can commit in a different order on each node; only when quiesced will the state become the same (i.e. the Kemme R(SI) paper)
> [3] If we architect an rdbms for current workloads and hardware, it looks quite different from traditional approaches (i.e. the Stonebreaker "complete rewrite" paper)
>
> Would that be a fair summary of your talk?
>
>

Fair, but without the jokes. The major thrust of the theory section is
that serializability is stronger than necessary to ensure consistency,
but also that consistency needs a more rigorous definition. The 25 year
fixation on serializability rather than consistency has lead to an
unnecessarily constrainted view of the range of database implementations.

Then, of course, there's an architectural description of NimbusDB,
albeit at a rather high level.


--
Jim Starkey
Founder, NimbusDB, Inc.
978 526-1376



[Non-text portions of this message have been removed]