Re: extension and interoperability

C. M. Sperberg-McQueen writes:

> This morning's call makes me think that some members of the CG believe
> we face a choice between writing the spec so as to have interoperable
> processors that do not extend the spec and writing it in a way that will
> lead to non-interoperable extensions.  I think this is a misconception.
>
> Like all spec writing groups, we face a three-way choice:
>
>   - write the spec in such a way that there is controlled extensibility
>     and controlled extension

>   - write the spec in such a way that there is uncontrolled extension

>   - write the spec in such a way that no one uses it

The first two of these items are badly phrased because they suggest that
every spec in use will be extended.  They should have said

    - write the spec in such a way as to offer clearly marked mechanisms
      for extension and with them the possibility that any extensions to
      the spec will be identifiable and controlled

    - write the spec in such a way as to offer no mechanisms for
      extension, guaranteeding that any extensions to the spec will be
      unidentifiable and uncontrolled

It is possible that a spec can be used and that no one will want to
extend it.  Designing it without any extension mechanisms is not a
reliable way of achieving that end.

-- 
C. M. Sperberg-McQueen
Black Mesa Technologies LLC
http://blackmesatech.com

Received on Tuesday, 1 February 2022 21:30:14 UTC