Re: Completeness

At 04:37 PM 9/28/2002, Mark Baker wrote:

>On Sat, Sep 28, 2002 at 08:19:49AM -0600, Champion, Mike wrote:
> > >  Show me a problem that Web services claim to
> > > solve that the Web doesn't have a solution for.
> >
> > The same can be said (proven?) about a Turing machine.  I can imagine how
> > the REST operations can be mapped onto Turings operations, or onto Codd's
> > relational algebra, thus I will believe from my little thought experiment
> > that you can solve any problem with the Web.  Your point is well taken.
> > But uhh, so what?
>So, if you and David agree with that, I don't see how it can be claimed
>that the Web is for humans.  If it's complete, then it's complete.
>This means that all tasks have at least one solution within the
>constraints of each architectural style.

All turing equivalent systems CAN be made to do the same thing. It's just a 
small matter of programming :-). The question is how hard is it. And how 
usable is the end result. In particular, not all computationally complete 
systems have the same performance, storage requirements, etc., properties, 
when applied to a given problem. Your only guarantee is that they will 
eventually compute the answer.

The way I interpret David's argument is that screen scraping html pages in 
order to fill in forms, while theoretically achievable by a computationally 
complete system is "too hard" to do today. We do have natural language 
understanding systems that can be made to do useful things in constrained 
worlds, but I've spent the last 20 years being in the state of having to 
wait another 3-5 yrs. (Always doing a get is not hard, but figuring out 
what the bag of bytes you get means so that a program can do something 
useful with them, is.)

>The BIG "So what?" here, is that this solution on the Web has the same
>properties that made the Web succeed, because it respects the
>constraints of the Web that induce those properties.  Since a Web
>services solution does not, because it does not follow some of those
>constraints (specifically, uniform interface) it will not share those
>same properties.

Logic foul!!!

p=respecting all the constraints
q=success (has the desired properties)
   p->q most assuredly does not imply ~p -> ~q (except in advertising:-)

The only way the argument could be correct if it is a tautology.
  p->q and ~p -> ~q
which is the same as saying p <-> q

  Success is achievable iff all the constraints are followed.
To my mind, this is just a restatement of "the one true way" argument and 
supportable only by the tautological assertion.

Please note: I am not saying that there is nothing useful to learned from 
REST. Quite the contrary. I am in complete agreement with David, and all 
the others who point out that there are many useful lessons to be learned 
from it. And it would be quite profitable to incorporate the ones which 
apply in a web services context. Just like it is quite profitable to 
incorporate the many other lessons that have been learned from 20+ years of 
distributed computing experience. (Heather pointed out several.)

I'm just tired of the dogma.


Received on Saturday, 28 September 2002 20:34:57 UTC