interface definitions (was Re: strong vs. weak typing)

Ken MacLeod <ken@bitsko.slc.ut.us> writes:

> I've done my first mid-level reading through the working drafts and
> I'm noticing a heavy bias towards strongly-typed applications and
> languages, but it's not stated in the goals whether this is
> intentional or not, outside the brief reference to CORBA, DCOM, and
> RMI in the introduction.

After a more thorough reading, strong/weak is probably the wrong way
to describe the issue I see.  The next descriptions I thought of were
well-defined or strict, but those aren't exactly right either.

The issue I see is that the working drafts seem to imply that HTTP-ng
requires interface definitions to be available to the protocol
implementations in all cases, there's no statement or implication of
support for a protocol layer that could work without explicit
interface definitions.  No opening for HTTP-ng-lite  :-)

This would seem to make HTTP-ng much more difficult to use for
prototyping, ad-hoc development, scripted applications in common
scripting languages, simple clients or servers, etc.

I'm still confident that the general architecture can support both
modes of operation, but I fear that if support for working without
explicit interface definitions isn't included as a goal, it may not be
elegantly retrofitted at a later time.

-- 
  Ken MacLeod
  ken@bitsko.slc.ut.us