xml:id RE: [sc34wg3] Compact syntax requirement question

Robert Barta sc34wg3@isotopicmaps.org
Sat, 23 Jul 2005 11:05:58 +1000


On Thu, Jul 21, 2005 at 06:39:07AM -0400, Patrick Durusau wrote:
> No, actually a much simplier point. Specifying the formal requirements 
> for a language, using EBNF syntax for example, could be a different task 
> than specifying a "particular" syntax in which the language was going to 
> be implemented.

"A requirement to use EBNF" to describe a TMQL, just means that there
is a selection what language I use to define a language (structure).
If the TMQL editors would just have written

  "We use EBNF, thanks for reading."

then you would not know ANYTHING how to actually write a TMQL
statement. Instead, the only reasonable thing to do was to
specify

   "[a] particular syntax in which the language is [written]"

"implemented" is no a good word here, because it has the connotation
of actually writing programs (eeeeh, the dirty word for the
conceptionalist, yes ;-).

> Obviously the authors of TMQL intended to produce a "plain text
> syntax" (and not a model for some _one_ arbitrary syntax) but it was
> not apparent from the draft.

Oh.

> The question arose here from a presumption of the authors that a plain 
> text syntax should be the norm. From the ensuing discussion, it appears 
> that presumption was not universally shared by all concerned.
> 
> As far as a plain text syntax being better, I note my point that neither 
> HTML nor XML have seen the need for an equivalent plain text syntax  has 
> gone unanswered.

Sorry. I guess there was no answer, because the argument seemed to be
off-target. XML is a (set of) notation(s) to organize textual content.
If an XML vocabulary (application) is defined it may have a meaning
(semantics), but most will not care, except the programmer who
actually makes something happen when you had typed <b>whohieee</b>.
What <b> actually means is loosely defined for HTML.

TMQL is (like XPath, ....) a data access language. One for a
particular way how the data is organized. Here things must have a
pretty strict form and meaning if we aim for interoperability.

Writing there "ah, and btw, what this exactly means" is not a viable
option, as is perfectly for XML applications. If I add a tag <rumsti>
to my HTML document, no browser goes up in smoke (well). If I add a
bogus character into a TMQL statement, hell should brake loose.

To bring it to a succus:

   "XML is a way to organize content into a tree form"
   "TM is a way to organize content into some-sort-of-graph form"

TMQL, XPath, SQL, etc. have a completely different objective.

> Not really. More of a failure to share the same presumptions about the 
> level of abstraction at which a draft is being written. The TMDM, for 
> example, specifies no syntax, which is covered by other parts of 13250. 

Mumble. On the surface, it appears so, but:

TMDM specifies how data (content) is organized in a "TM way". So it
defines how a particular structure is to be built. It uses
_structural_ descriptions based on the Infoset (not on, say, EBNF).

So it is also 'syntax', not one for textual organisation, like
XML/XTM, but a syntax for a multidimensional thing (of course, TMDM
adds more constraints to it, so it is "not just another syntax").

\rho