Re: Question on File locking

On 16-Aug-2014 06:36 -0500, D*B wrote:

On 14-Aug-2014 08:10 -0500, CRPence wrote:

<<SNIP>> Even if a timestamp with necessary precision is not
guaranteed to be unique, there is of course, for example, the
result of the SQL scalar function GENERATE_UNIQUE() [generates a
UUID] which is a unique value satisfactory for sequencing and that
includes a TIMESTAMP. <<SNIP>>

... a couple of years ago I've evaluated GENUUID on AS/400; the
documentation ended up with "most likely unique" and testing in a
heavy parallel environment brought up collisions; as far as I
remember the implementation did not include the processor number,
with the result of possible duplicate values on multi processor
boxes.

A result that is a clear contradiction both to the documentation of the feature and to the /standard/ effect, is IMO, clearly a *defect* that should be reported. Given the tests were conducted so long ago, there is a very good chance the feature has been corrected, because someone who had recognized the contradiction cared enough about proper results [for themselves and possibly for others] to report the issue.