Thomas Lockhart wrote:
>
> > SQL'99 explicitly forbids it. Please refer to my posting to HACKERS
> > for the SQL document reference.
>
> The fact that a standard "forbids" something does not necessarily mean
> it is a bad idea, as I'm sure you know. Is there any reason that the
> standard forbids using domains inside arrays, other than someone on the
> standards committee realized that it would be hard for their company to
> implement it? That is, does allowing domains in arrays lead to
> inconsistancies or fundamental issues with relational algebra or other
> set logic that should keep it out of the next set of standards?
>
I partially agree, but I guess Tom has already given some of the reasons
not to do it.
> If Postgres was developed to only the current standard, it would never
> have been written. And since the start of the open source days, if we
> had worked solely to get it to conform to the current standard we'd be
> starting at ground zero for implementing SQL99, since many of our
> features now appear in that standard. Someone cheated and looked at what
> we could already do... ;)
>
Again, I only partially agree with that, Adding significant features
that
will allow people to solve significantly different problems that can not
be solved with the vanilla standard is a good think. And I believe it
is
acknowledged in many places that many SQL3 features were inspired on
Postgres.
However, adding extensions to the SQL standard otherwise is a bad thing.
If affects portability. Actually, "extending" standards has been a
weapon
used by some proprietary companies to hurt the competition. Standards
are
friends of Open Source software and we should try to stick to them
whenever possible.
In the case of DOMAINS, which are already considered by some as not very
useful and passive of removal from next editions of the standard (by one
SQL editor, at least -- I can give you the book reference this
afternoon),
adding extension to the SQL to it would just aggravate the issue.
So, although I agree with you in principle, I believe that in these
cases
we should stick to the standard and avoid gratuitous extensions.
--
Fernando Nasser
Red Hat Canada Ltd. E-Mail: fnasser@redhat.com
2323 Yonge Street, Suite #300
Toronto, Ontario M4P 2C9