[cvsnt] Re: Repository auditing with Oracle

[Tony Hoyle]

Glen Starrett wrote:
> I usually resort to using a GUID for multiple-location instances where 
> ID just wouldn't work.  It would be helpful in this case too, since the 
> GUID could be generated by CVSNT independent of which DB is running.

The proglem with GUID is the slowdown from passing them around the 
program.. integers are nice and simple for that.  That's quite a 
significant difference in 2.6 (especially once all the caching is in... 
map lookups on integers are really quick).

It's not really an issue for the audit of course.

> The downside is a less readable PK and larger field size.  Overall 
> though it would probably be the more 'proper' way to approach it with a 
> DB layer that contains the necessary DB-specific ID generation step for 
> Oracle.  Of course that's difficult to say as a blanket statement -- I 
> don't know the requirements that are used as the basis for setting up 
> the schema.  I think though when I read between the lines, that is 
> exactly what has been done.

Pretty much, yes.. that part of SQL has never been standardised anyway.

Tony