Re: Genericity, strong typing, SOAP-RPC, GRASP, and defining the Web

[taking the discussion offlist after this one, fyi.]

Miles Sabin wrote:

> It's often the case that architectural
> simplicity comes at the cost of implementation complexity, in which
> case measuring complexity by focussing soley on it's architectural
> aspect is cheating.

This is (again) a common objection, but it doesn't necessarily map to
reality.  It's important when looking at systems not to assume that they
have an optimal level of complexity as implemented;  many, many systems
introduce unnecessary complexity of both design and implementation ---
hey, we engineers are humans too, right?  So perhaps when evaluating
systems it's useful to distinguish between unnecessary or artificial or
inadvertently introduced complexity and the minimum complexity inherent in
the problem domain.  Given the high degree of complexity inherent in
something like an operating system, I'd suggest e.g.  Plan 9 stands as a
great example of architectural simplicity *leading to* implementation
simplicity.  I don't think it's "cheating" to point out existence proofs!
:-)

> I agree up to a point. But this only applies where HTTP is used
> solely for it's original purpose: hypertext transfer. Add media types
> with type specific processing at the receiver, or content negotiation,
> or layered application semantics encoded into URI conventions and the
> genericity evaporates.

I'm not so sure.  It becomes less obvious for sure, and complexity does
increase.

jb

Received on Tuesday, 26 March 2002 21:35:27 UTC