[Monotone-devel] Re: Proposal for human readable revision IDs

[Bruce Stephens]

Florian Weimer <address@hidden> writes:

> * Bruce Stephens:
>
>> Many SQL databases give each row a UID; it may be that SQlite does
>> that.  Ah, it does, I think.  I'd guess the column ROWID, OID, or
>> _ROWID_ of the revisions table would provide roughly what you want?
>
> Would this ID survive schema migration?

I doubt it.  I don't know how such migrations are usually done; I'd
guess for some, it would survive, and for others, it might not (which
might be worse than having it always not survive, I guess).

> Probably not, and this makes it a bit too volatile even for local
> revision numbers, doesn't it?

Does it?  That doesn't seem much of a problem to me (for one thing,
schema migrations are presumably expected to be rare), but it would
all depend on what you wanted to do with these revision numbers, and
what properties you expected them to have.