Posted by Joe Weinstein on 02/06/06 01:11

Erland Sommarskog wrote:

> Joe Weinstein (joeNOSPAM@bea.com) writes:
>
>>Ok. Thanks. However, by default, with SQL2005, why is this a deadlock?
>
>
> I didn't touch that part, as it was already explained in other post, but
> in case you missed it.
>
> First both processes inserts a row into the table, and maintain a lock
> as the INSERTs are not committed. Next, they try to read the entire table.
> But do that, session 1 needs to access the row inserted by session 2
> and vice versa.

Gotcha. And if it were 'snapshot' isolation, each read would only see
their own row, I suppose, and continue.
thanks
Joe

[Back to original message]