On 6/2/2007 9:28 AM, Andrew Sullivan wrote:
> On Fri, Jun 01, 2007 at 06:27:12PM -0400, Christopher Browne wrote:
>> I have noticed a scenario (whilst running DDL tests) which causes me
>> to want to add an extra test to the pre-processing of this...
> 
> Can we not add this to .10, and make this a KNOWN ISSUE when we ship? 
> This is a corner case, and I want to work out exactly the
> implications of your proposed change before accepting it: we've
> gradually been expanding the locking behaviour of various parts of
> the DDL handling, and it's starting to make the system very
> inflexible.

I've been thinking about this a bit more and the proposed changes don't 
solve the problem at all, they just make the window in which this race 
condition can occur smaller.

The DDL script is injected into the system at the origin of a set and 
flows like the data out to the subscribers over their prividers. The 
subscribe however is injected at the new subscriber, travels to the 
origin which will respond with an ENABLE_SUBSCRIPTION that travels back. 
At the time the SUBSCRIBE_SET is created on the subscriber by slonik, 
the subscriber might be behind, thus hasn't processed the outstanding 
DDL script event yet and would just happily accept the subscription request.


Jan

-- 
#======================================================================#
# It's easier to get forgiveness for being wrong than for being right. #
# Let's break this rule - forgive me.                                  #
#================================================== JanWieck at Yahoo.com #