Lets have an overview of different proposals and see what we prefer:
My proposal for this specific thing is:
Use only conventional record schemas in the recordSchema parameter,
concatenate them  in a fixed  container <srw:container> and do not add
the complexity of other envelopes.


>>> [log in to unmask] 11-08-2005 15:51 >>>
> Date: Thu, 11 Aug 2005 09:40:37 -0400
> From: "Ray Denenberg, Library of Congress" <[log in to unmask]>
> From: "Robert Sanderson" <[log in to unmask]>
> > We could change the semantics to:
> > [...]
> I still feel strongly about overloading a parameter, which is why I
> suggested a new parameter altogether.  The problem with that is we
> would have to wait until the next version.

I agree that overloading parameters is bad.

However --
(A) there is no parameter-overloading in my suggestion of using URIs
    constructed according to some documented scheme;
(B) we have a well-defined extension mechanism so there would in any
    case be no reason to wait until v2.

/o ) \/  Mike Taylor  <[log in to unmask]>
)_v__/\  "There is a huge switching cost to using a different
         system.  It is this switching cost that has given customers
         the patience to stick with Windows through all our mistakes,
         our buggy drivers, our high TCO, our lack of a sexy version"
         Microsoft's C++ Manager Aaron Contorer, in a memo to Bill