>I've tried both methods during 15 years of programming, and found the 'prefix' method is easier to handle. Like the range it allows you to see where the data was first input, but it never has to be updated if one site fills up its quota of ID codes.
>
>My $.02 US
>Barbara

I've ended doing the same, using two char code as the origin marker, and make the rest with sys(2015) (or so, I'm weak with remembering names). It's FPD so I'm using character key. This is great for an app which will be promoted into distributed processing this harvest (yes, it's agricultural app). Some records will be generated on the remote installations, and some will have to be sent back after some processing. Knowing the originator or knowing the customer's preferred originator will greatly help when I need to select the data to be sent back to the remote locations.

In terms of integers, I'd probably use some tiny numeric code to denote the originator, shift it all the way left, and do BitOr with the generated key (provided the generated key never uses these higher bits).