Grant Fritchey (5/31/2012)
...I was trying to make the point that "generic" database designs aren't really possible due to the nature of the underlying infrastructure. Clearly failed.
Well you didn't completely fail because that was my best guess. (I reread the article but still wasn't sure - hence my post)
It is a good point about the foundation of each.
I hadn't thought about the foundation of SQL Server being clustered vs. the foundation of Oracle being the heap. This is an interesting discussion (though the topic has digressed).
I learned on Oracle in `92 while still in college, and I have used it often since then, but I haven't done a db design in Oracle... uh, this century?
So, my follow up question is:
Can't you still write specific optimized code for both Oracle and SSvr in one program if you have two different DALs with different schemas underneath?
Yes, you'd need a forked code base for the DALs, but if you're dealing with 'big data' ("big enough data"?) and tight performance constraints then it could be worth the headaches and (huge) costs.
Maybe that's a dumb question, forgive me as I'm a applications developer (by job category) not a a DBA - though I do all the SSvr db design and query optimizing for my area 🙂
I do now agree with your premise which I think I can over simplify as "Don't gold plate your code" (the gold plating in this case being the fiction of the generic backend database)
edited to add "in one program" to the DAL question