Print

Print


You could declare it to be mandatory, but the implementors will cope with it
being absent, as it is typically unnecessary.

Ralph

-----Original Message-----
From: Ray Denenberg [mailto:[log in to unmask]]
Sent: Wednesday, July 10, 2002 4:53 PM
To: [log in to unmask]
Subject: Re: multiple schemas


"LeVan,Ralph" wrote:

> But (a) is not amenable to explain.  It says that the behavior is
> unpredictable, learn to like it.  I don't.

(b) also says the behavior is unpredictable. The only difference is
that you'll get all records in the same schema rather than possibly
different schemas, but you don't know what schema, so it's just as
unpredictable.

If it's unpredictability we're trying to avoid, I'm left to infer
that the semantics of omiting the schema parameter is that the client
knows the default schema, and wants that schema.  In other words,
there's always a schema requested, implicit or explicit.

In that case, let's just make the schema parameter mandatory and
avoid this confusion.

--Ray



>
> Ralph
>
> > -----Original Message-----
> > From: Ray Denenberg [mailto:[log in to unmask]]
> > Sent: Wednesday, July 10, 2002 10:46 AM
> > To: [log in to unmask]
> > Subject: multiple schemas
> >
> >
> > "LeVan,Ralph" wrote:
> >
> > > I'm unhappy with the opinion that not specifying a schema
> > might result in
> > > records from multiple schemas being returned.  Servers
> > should be expected to
> > > specify their default schema (through explain) and return
> > all records in
> > > that schema when an explicit schema has not been specified.
> >
> > What are the semantics of omitting the schema name in a
> > request? Is it:
> >
> > (a) give me each record in whatever schema is available (or
> > the best, if there
> > is more than one), or;
> > (b) I don't know what's the default schema but give me all
> > records in that
> > schema; or
> > (c) I know what your default schema is; I'm omitting it
> > because I'm lazy. But I
> > want all the records in that schema.
> >
> > If it's (c) then you're right.  (b) doesn't make sense -- it
> > assumes that the
> > client is prepared for multiple schemas so why limit it to
> > one.  If it's (a)
> > then I dissagree.
> >
> > --Ray
> >