ON STANDARD COMPATIBILITY VS. RELATIONAL FIDELITY
with Fabian Pascal

 

 

 

From: Fabian Pascal

To: Tom Kyte

Date: Jul 11 2005

 

What exactly did you mean by this?

 

Yeah, but saying 'It's standards-compliant' is like saying "my HR system uses the 42nd normal form, primary keys, foreign keys and has zero NULLABLE columns". In a way :)

 

 

From: Tom Kyte

 

The intent was to say in a tongue in cheek way that is like saying 'my stuff is totally good because it is standards based' ... My stuff is good because it is built on open standards. Not my stuff is good because it works, is the best at what it does.

 

Just and offhand comment on a blog, nothing overtly technical or anything.  A comment made after seeing one too many systems "architected to be the most open thing, built using every standard and buzzword. Yes it is unfortunate they don't actually work or perform their intended functions, but aren't they a thing to behold." It annoyed me that his fully standard compliant site didn't work with my browser that morning.  End users don't really care whose fault it is, they do want things to work.

 

 

From: Fabian Pascal

 

I don't have a problem with that argument, but your comparison—"is like saying 'my HR system uses the 42nd normal form, primary keys, foreign keys and has zero NULLABLE columns”—can mislead in an ignorant industry.

 

Fully normalized, fully constrained, no-NULLs databases are not the same as standard compatible databases. The former means soundness, a requirement of good design, which makes them work; standards can be arbitrary and unsound (witness SQL), so even compatibility with them does not, indeed, guarantee that they work.

 

Besides, there are two different kinds of "good" and "works" which should not be equated. The design of an application (including a web site), is not the same as the design of a database; the former does not have a theoretically sound basis, the latter does.

 

I would also refrain from expressions such as "42nd normal form" for reasons that should be obvious.

 

 

Posted 9/9/05