[BioSQL-l] Pg and mysql versions are decoupled

[Hilmar Lapp]

On Wednesday, March 19, 2003, at 03:15  AM, Matthew Pocock wrote:

> Is it easier to go mysql->pg or pg->mysql, or something_sane -> both?
>
> M
>

I guess it is always easier to go <something richer>-><something less 
rich>, so pg->mysql is supposedly easier.

There are some XML definitions for declaring relational schemas out 
there, which would be totally RDBMS neutral. An example is Torque 
(jakarta project), which would provide the DDL renderer along with it 
right away.

I'd personally prefer going that (XML) route over the long term, even 
though ChrisM has argued in the past that people can generally and 
quickly make good sense of CREATE TABLE statements, and that SQL/DDL is 
expressive enough. I agree with this but my standpoint is having an 
upfront XML definition which a renderer converts into your favorite 
RDBMS's DDL statements doesn't mean you're unable to look at and 
inspect DDL statements. It's just for changing them that you'd have to 
go into the XML layer.

I used to have reservations against an XML meta definition because then 
you couldn't use ERD modeling tools to auto-generate the DDL anymore 
(because none of them generates XML - even though with a decent amount 
of digging around and an expensive license you could probably get 
template-driven generation tools like ERwin to achieve this). However, 
it turns out that over the long run the rate of change diminishes and 
can be easily maintained by hand, and, more importantly for an open 
source project, there are no good and free ERD modeling tools around 
which also let you auto-generate DDL for different RDBMSs.

	-hilmar
-- 
-------------------------------------------------------------
Hilmar Lapp                            email: lapp at gnf.org
GNF, San Diego, Ca. 92121              phone: +1-858-812-1757
-------------------------------------------------------------