Date index for Aug 2002

Hi,

Dave Nuttall wrote:

> However, a couple of points for the sake of discussion:
> 1) Don't we have a basic mixture of concepts where the usual idea of an 
> AUTO_INCREMENT (or IDENTITY) field is maintained automatically by the 
> RDBMS system and not artificially controlled as in the current Achievo 
> mechanisms?

Yes. I agree this is a mixture. But is is for the sake of database 
interoperability. Most RDBMS's don't have an auto_increment feature on 
field level. Most of them use sequences (for example Oracle, PostgreSQL).
Achievo's db_sequence table is more equivalent to sequences than mysqls 
own auto_increment feature.

> 2) Does the size of the varchar field in db_sequence pose any possible 
> problem in cases like mine where I happen to take the route using very 
> LONG table names?  (i.e. my Technlogy_SubClasses exceeded the field 
> length).

Yes, it could lead to problems for very long node names. We can make the 
field a bit longer if necessary. Or, cleaner, but timeconsuming and HELL 
for backwardcompatibility: make the database abstractionlayer use mysql 
auto_increment if possible.

There might be a workaround for using mysqls auto_increment feature, but 
that's untested and I'm making it up right now, but it might work:

leave out the AF_AUTOINCREMENT flag, and instead set an AF_HIDE_ADD 
flag. Not sure if that will work, but that might prevent Achievo from 
generating an id.

Greetings,
Ivo