RE: [xml-dev] SGML complexity

[sterling <sstouden@thelinks.com>]

Basics of English is learned by the 2nd or third grade, but most are in 
college before they can publish with it.  It takes practice and more 
practice and it took several centuries for it to evolve into 
a standard.  SGML is, I agree, not too difficult.  

The higher level tools for SGML have not been worked on by a large 
community of developers, presumably because as you say a language that 
serves everyone is too broad or as I originally thought, because the SGML 
was destined to serve a specialized pre Internet community of users. 

I just wonder, if developers returned to the root and started over what 
could be done. 

What exactly do you mean "... attempts at universality reflects the 
complexity of the task as much as the tool"?  

Tools should resolve complexity to functional utility, but without 
complexity there would be little chance to achieve universality. 
Differential equations and vectors are tools that handle the complexity 
of real world space, time and shape problems. Few understand the tools, 
fewer still understand the problem, and no one has a complete grasp of the 
whole, still it works. Reduce the complexity of the universe below its 
point of universality and you lose the ability for it to support its 
current dependents.  

The deficiency is "tools that make useful" the complex.

To restate the problem as a need:

A language comprehensive to all information.
sterling


On Fri, 8 Sep 2006, Len Bullard wrote:

> It doesn't take that long to learn the basics of SGML, and even the advanced
> bits aren't too hard to use.  It helps if one has some familiarity with
> computer science before beginning, or at least are comfortable with
> notations such as BNF.  It depends on the task:  tagging comes naturally to
> most people with a bit of practice; DTD design requires a good abstract
> mind, and so on.  The "SGML Is TOO HARD!" rubric is oversold and a bit
> whiny.
> 
> It is implementing SGML software itself that is complex.  It has lots of
> features and therefore, many corner cases that one won't discover until one
> is deep into the muck.  This like so many attempts at universality reflects
> the complexity of the task as much as the tool.  Attempting to have one
> metalanguage that works in all languages, syntaxes, and glyph systems is a
> like stretching a cloth over a forest of forests and expecting a cheap,
> tight fit.  It ain't happening.
> 
> len
> 
> 
> From: sterling [mailto:sstouden@thelinks.com] 
> 
> Yes, the short term test is "community" but in the end, the technology has 
> to be irreversibly satisfying to some market need and adopted by all of 
> the user members of that market.  The products that support and enable 
> must satisfy the essence of that need. 
> 
> As I see the market need, it is universal, not niche. The real market is 
> for a universal solution capable of solving any demand for ordered data 
> presentations.  Few want to learn some of the 20000000000000000000 
> different --ml languages for each segment of it. 
> 
> Most want a language like English, that can be used to present and 
> express anything that comes up.  SGML does that.   It takes about 15 years 
> for  members of the general population to learn the English language, how 
> long  would it  take for the general population to learn SGML?  English is 
> one of the most hated sujects in grammar and high school and I think SGML 
> would be too, but it is the only language that could be truly universal.
> 
> I think the targeted community approach will work for a while, but in 
> fewer than six or seven years you might run out of targeted 
> niche communities.  There are still a few commodore computer fans!
> 
>