Kin Recognition: Why XML 1.0 Is Core and Everything Else is Optional

["Bullard, Claude L (Len)" <clbullar@ingr.com>]

Matthew Gertner:

>Well, the issue we are discussing is not so much whether or not we need
>another schema language as whether we need *any* schema language.

>If you're talking about defining standard vocabularies or "core components"
>then I agree, that's a more pertinent question. Without that, none of the
>web services or semantic web business is going to fly.

This is the environmentally acquired labels vs genetically 
acquired labels problem of kin recognition.  One needs to 
do both depending on the homogeneity or heterogeneity of the 
habitat.  What kinds of cue selections favor opportunities 
to establish membership?  (Why is the WSIO the fastest 
growing consortium in recent history?)

http://www.bio.unc.edu/faculty/pfennig/lab/pfennig_files/kinrecoginization.htm

"Generally, the relative importance of genetically-based versus environmentally-acquired labels depends on which type of cue is less likely to cause a recognition "error."  Relying solely on environmentally-acquired cues might cause individuals to mistakenly assist nonrelatives that inhabit the same environment.  These nonrelatives could reap the fitness rewards of beneficence without reciprocating, and so increase in frequency in the population.  Relying solely on gene products also might cause individuals to mistakenly assist nonrelatives that carry "outlaw alleles" - alleles that encode the recognition cue, and which spread by helping themselves, but not the rest of their bearer's genome.

The likelihood that these errors will occur depends on each species' genetic system and ecology.  Organisms can minimize the chance that two nonrelatives will share similar gene products by using cues encoded in highly polymorphic genetic loci, such as MHC loci.  Such genetically-based labels are most useful for organisms that occur in relatively homogeneous environments.  For organisms that occur in more heterogeneous habitats, environmentally-acquired labels permit accurate discrimination of kin without the potential risks of outlaw alleles."

Why RPC?  Other than the utility of SOAP formats, WSDL, and maybe UDDI, 
if one uses RPC, the need to build up all the standard language 
constructs (eg, the Industrial Schema) go away.  All agreements 
are local and come down to kin recognition and eliminating false 
correlates (outlaws).  It's based on self-selection among 
members.  Google may succeed precisely because of abandoning 
the principles of the web architecture and instead pursuing a 
selfish strategy that enforces kin recognition through learned 
cues (nature favors lazy selections) that build up community 
membership.  IOW, go directly against the anytime anywhere 
any information selection and choose a strategy of membership.
Lock in, pure and simple but is reduces exposure to semantic 
drift to a simple set of interfaces rather than the highly 
probable differences of semantic interpretation of the schema 
from hell.

Maybe the problems are the big committees and the industrial 
agreements if the ecosystem itself favors local kin recognition 
and laissez faire breeding (not too close, not too far).

Nah!  You have to be able to do both and you have to choose. 
XML can't favor one or the other within the core.

len