in reply to Re: Sequences, last_insert_id, SQLite, and Oracle
in thread Sequences, last_insert_id, SQLite, and Oracle

Thanks — this is an interesting approach I hadn't thought of.

My IDs should ideally be fit for human consumption, so 128 bits is a little long, though. The kind of volume I'm expecting, there won't be more than a few IDs to generate every hour in phase one, then maybe a few a minute in phase two, but with rather fewer than that on average. In terms of address space, an eight- or nine-numeral wide decimal number ought to suffice for the lifetime of the project and that allows for quite some margin of error. Of course I'll allow for more than that in the database, but in thinking of how this would be used I'm comfortable with assuming the user would occasionally (not often) have to handle an identifier like "6126-7245".

  • Comment on Re^2: Sequences, last_insert_id, SQLite, and Oracle