Re: Dictionaries in the library

> Re: Dictionaries in the library
> 
> From: Tim Berners-Lee (timbl@w3.org)
 
> With formally specified languages, the spec is
> in 1:1 correspondence with the language. 

Not always: a formally-specified language could also be specified with
a juridical process to guide interpretation and overcome flaws. (e.g.,
the errata process for XML).

It could also be specified accompanied by a "community tradition"
process
as well, saying that if most/all implementations interpret an ambiguous
text 
one way, that that is its meaning.

(What is missing from most specs is a statement "please if you find an
ambiguity,
choose the less crazy interpretation", which may successfully divert
legalists
onto edifying dissections of craziness. :-)

It is difficult to specify formally anything using just text: if you use
an
artificial language no-one understands it and errors creep in; if you
use 
technical English, it is not very expressive and prone to mistakes. This
is
why I think one must consider standards-reading as ultimately a social
activity based on agreeing on respect/competency/experience/power roles;
even artificial languages are human activities.

To say that a language is defined by a formalism will, when that
language
gets used by communities, immediately fall down: few formalisms are 
expressive enough to specify the software engineering intentions
underlying
its design. When SGML uses the term "generic identifier" it embodies a
lot which 
cannot be expressed in a high-level language definition system such as
Z, for example. 
But "generic identifier" is a concept at the heart of SGML as a
language.

Rick Jelliffe

Received on Tuesday, 23 May 2000 11:43:27 UTC