[
Lists Home |
Date Index |
Thread Index
]
Ian Graham <igraham@ic-unix.ic.utoronto.ca> writes:
>
> On Tue, 13 Jan 2004, Bullard, Claude L (Len) wrote:
>
> > A troll of sorts but with a serious core:
> >
>
> A mathematical system can certainly be noise free and
> complex. Also adding noise to a system that is not complex
> can make it so.
Yes, but such a mathematical system is a purely a theoretical model, no
real physical system could be implemented that would be noise free?
> However, both statements imply you have a completely
> specified mathematical model for the system, within which you
> can test for complex behavior. I would argue this is not true
> for XML, since the 'rules' of the model are still changing,
> so the point is (in the long term) moot.
>
> On the shorter term, a question might be: given the current
> XML model, are there some properties of XML that are complex
> (with some definition of complexity, such as divergence of
> semantics), or that becomes complex with some change in a
> model parameter (such as system size). I think the answer
> might be yes, but I don't think anyone's proven it.
I'm really surprised no one took me to task for my earlier suggestion
that Shannon would have said "no" to Len's questions? Perhaps because
there really are two separate issues here?
The first question -- the one I answered previously -- whether a system
can be noise free really isn't relevant in my mind. I think the real
question is whether a system can be robust enough to not be modified in
any significant manner by any noise that it might encounter? Clearly,
the answer here is yes. Shannon proves that given enough bits (atoms,
whatever) you can do this, though equally clearly, "enough" might not be
a practical number...
To directly address Len's 2nd question, what we need to add to XML is
error detection and correction. Given the Schema mess this could quickly
degenerate into a discussion on what errors do you care about? However,
given the way Len poses the 2nd question I have to wonder why MD5 hashes
or digital signatures are not sufficient to pick up the fact that the
XML has changed?
>
> A second question would be: can one prove that the 'model
> space' can evolve such that the system transitions to
> non-complex behavior, or has scaling behavior that is
> different (i.e. changes such that it takes a 'larger' system
> before things go complex). I've no idea how you would prove
> that, but it sure is an interesting question!
Again, I think Shannon shows how this might work: add more bits (that
provide the necessary robustness)?
> Ian
> > Is this possible:
> >
> > Can a system or a system of systems be both
> > complex and noise free? What must be true
> > of all of the components of the system for
> > this to be the case? By what properties
> > can one measure this? Are there complex
> > systems that can not under any circumstances
> > be noise free (beyond XML-Dev itself before
> > the deconstructionists flame on)?
> >
> > Why I'm asking:
> >
> > This is related to Berners-Lee's recent
> > post to the TAG list on the requirement
> > for equality of 'chunks' of XML regardless
> > of the system component handling the
> > chunk. In other words, what is needed
> > for any XML part or whole to remain
> > invariant across any current or future
> > WWW system (the WWW is a system of
> > systems)?
> >
> > We should consider this problem on XML-Dev.
> >
> > len
> >
|
