RE: [xml-dev] The Rule of Least Power - does it miss the point?

["Bullard, Claude L \(Len\)" <len.bullard@intergraph.com>]

Not just loose coupling, but variation given 
repetitive inputs, or novelty seeking.  Thus, 
filtered couplings and what Pangaro called 
'through looping'.

Today credit systems are pattern seeking. 
They aren't always precise and can turn 
off your credit without you doing anything 
wrong, but generally, they do detect anomalous 
activity and only err to the conservative side 
somewhat like the problems of SmartFilter and 
content and the challenge of filtering porn 
from kids and adults (generally, the kids will 
get it and the adults won't but never mind).
Any smart mammal that detects filtering can 
game it so any pattern seeking system has 
to be a learning system, thus, pragmatics.

That maximum indexability is orthogonal 
to learning is a red herring.  It is a 
so but so what.  You still have to filter 
given a system that maximizes opportunism 
without a knowledge of type: that is witless.  

The idea that this leads to a global filter 
is similarly wrong.  It leads to types of 
filters that are tuned locally.  There doesn't 
have to be a locus of control. In fact, that 
doesn't work.  It has to be lots of them 
that negotiate when they beat against one 
another unproductively.  Humans are also 
controllers.  The idea of observer as a separate 
system is somewhat bogus because the 'web' 
is as Fielding says, a social system and 
geeks have to deal with even if they are 
by design, anti-social components. ;-)

In response to Andrew:  since it is possible 
that interacting with my accountant was the 
source of the identity theft (a low paid 
person on her staff was jailed for fraud), 
of the options, TurboTax is the best risk 
given cost, knowledge necessary, and convenience.

I don't find embedded scripts abhorrent.  I 
find the idea that all pages must be equally 
indexable abhorrent.  Since page indexing was 
the use case provide in the RLP finding, it is 
the best place to begin a critique of the rule.

len

-----Original Message-----
From: Gavin Thomas Nicol [mailto:gtn@rbii.com]
Sent: Friday, March 10, 2006 10:21 AM
To: XML Developers List
Subject: Re: [xml-dev] The Rule of Least Power - does it miss the point?



On Mar 10, 2006, at 9:22 AM, Bullard, Claude L ((Len)) wrote:

> True for the global network.  False for particular
> information flowing at particular times in particular
> contexts.

"Particular context" sounds to me very much like a frame of  
evaluation == local processing. Even in closed systems involving only  
a few machine, only one system at a given time has a complete  
understanding of it's local processing context. The degree of  
variance in interpretation will likely be less than in open systems  
however.

I think it's fairly well understood that in general, loose coupling,  
is a good thing (not always, but usually). I think this "rule" (which  
I agree is a bogus term for it) is just a common sense design  
principal that is similar in nature to the principle of loose  
coupling... basically, it just says, wherever possible, make it easy  
to process the data you generate, and declare behaviours in a way  
amenable to analysis. Something similar can be found in "The Art of  
Programming" (I think chapter 7, might be wrong), where they say  
"encode your data in text wherever you can"...

Maybe I'm just of of the few people that finds embedded scripts, and  
things like DHTML, abhorrent for the most part.

> But the lesson is
> that unconstrainable use of a global identifier in a
> system that is quick to grant privileges and slow to
> revoke them is a bad thing.

This is really orthogonal to the issue of maximising the ability to  
analyse and reuse data.

> Another way to think about this:  if data is marked
> correctly with regards as to operations that can
> be performed on it, then the object can acquire
> rights from the governing environment or have them
> revoked.

Controlling capabilities is again, orthogonal to the ability to  
analyse/reuse data because it implies some global means of control  
that may or may not exist. You can declare what is valid on the  
object as much as you like, but ultimately, I will decide what to do  
with the bits on my box, unless we are bound by the same controlling  
mechanism. The likelihood of that every occurring is so low that  
you'd best keep your skeletons in your own closet, with the door  
firmly locked.