On Thu, Feb 23, 2017 at 4:08 PM, Pantelis Theodosiou <ypercube@gmail.com> wrote:
> Question: Will the patch be removed if and when Oracle decides to be
> compatible with the standard and forbids non-aliased derived tables?
>
> (I know it's a rather theoretical question. Unlikely that Oracle breaks
> backwards compatibility on that.)
Even if they did, so what?
First of all, our project's aim is not to copy Oracle slavishly but to
build a good database. Sometimes that involves making things work in
ways similar to Oracle and sometimes it doesn't. For example, I have
no urge to get rid of transactional DDL just because Oracle doesn't
have it. I have no feeling that NULL should behave in the completely
unprincipled way that it does in Oracle. And I don't think that
PostGIS needs to try to go be more like Oracle Spatial.
Secondly, extensions to the standard that let reasonable things work
which the standard doesn't permit are generally a good idea. We don't
want to let things work that really deserve to fail - for example
because the meaning is ambiguous - nor do we want to implement
standard syntax with non-standard semantics. However, neither of
those problems exists for this case. I don't see the point in making
things fail that could just as easily do what was wanted; that seems
pedantic. I don't think it's only Oracle that allows omitting the
alias; I think there are a number of other systems that behave
similarly.
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company