>Thank you all for your answers, I do use a key generating function, basically for every table, the key is a character string, I think it's easier to manipulate them in relations and that sort of thing.
>
>I also use table buffering and an error handler to take care of conflicts, the key generation part has worked well so far.
>
>**** Actual question here
>
>Basically I was trying to decide if I should stick to this double Id thing, mine and the user's, or should I go with the user's input, and validate it on creation, then use RI rules to handle changes.
>
>
>So far I've stayed away from RI rules, when I began programming in VFP I thought it could really slow things down if you change a part number, to go and update possibly thousands of records that contain the old value, I hadn't tried it though. The unique Id allows me to stay clear of this.
>
>My key generation program is also called newid(), a friend pased it to me, and I know he got it from someone else :), I've made some changes to it, among them, I took it out of the DBC procedures into it's own prg.
>
>Apparently the system assigned unique ID thing is the way to go.
>
>Thanx

Oh, slightly different question. In terms of carrying two id's ... If the customer wants it you pretty much need to do it. (Okay, if they really want a framitz, build them a framitz).

We have to deal with a mainframe in the mix of things here. Their data structures are a bit difficult to work with, and for some reason they really want business meaning in things like purchase order numbers. Any way, because the po# is one hard to construct, and two has a tendency to loop around generating repeating po#'s, we carry an integer primary key, and build their po#. When we run out of po#'s during the course of the day, we start replacing them with (*'s). Lately every time someone wants to know what a po# is they look for the primary key ... Change is possible, sometimes it's just harder than others. (But I digress)

As for using a character instead of an integer, you can do that, but there are a couple of reasons for using integers. One, they take up much less space in terms of bytes than characters. Two, they are much easier to increment.

Good luck,

Regards,

Jason