Hi Charlie

I think I'd prefer to live with having to occasionally synchronize the PK table against the actual tables. I assume the table header will be used for storage of the last generated PK. During a power outage, it will still be possible for the last generated key to become out of sync with the last actual key. So there will be a need to synchronize the two.

>Hi Dragan,
>Good to meet you in Kansas City.
>When I do a MAX(PK) SQL statement with a network table of 250,000 records, it takes 9 seconds. That's quite a penalty for each time a user starts up.
>
>>It is, simply because it's done low-level, somewhere in the table's header or so, so it's in a block which is already read, already has the locking mechanisms in place and generally does all the neat homework under the hood, and you don't have to worry about the stuff from the above paragraphs. While I may think that my newid() is nearly perfect (or let's say sufficiently perfect), I still know that that doesn't work perfectly in all of the situations. Native support is always better.
>
>Said sufficiently perfect.