Re: Pragma Proposal ready for review

> On 14,Dec2021, at 4:18 AM, Steven Pemberton <Steven.Pemberton@cwi.nl> wrote:
> 
> ...
> 
> I was expecting a proposal for a syntax to communicate with implementations, that further had no defined semantic content; something that allows things along the lines of
> 
>    {! ignore ambiguity}
>    {! serialize all parses}
>    {! serialize to json}
> 
> (though without specifying the "ignore ambiguity" bits, which would be specified by implementations).
> 
> Any other issue, particular semantic ones, such as namespaces, text insertion and so on, are separate use cases, that need to be discussed separately, but have absolutely no place in pragmas, because pragmas are exactly about things that are not standardised.

In the current state of ixml, namespaces, text insertion, and so on
are not standardized.

If someone wanted to support them in an implementation without
interfering with the operation of other processors, they could (1)
extend the syntax of ixml, (2) add whatever they need in comments that
have a particular structure, or (3) use pragmas.

Approaches (1) and (2) fail because they do interfere with other
processors: (1) encourages the users of the implementation to use a
syntax not supported by other ixml tools, (2) risks collisions with
other extensions which also use structured comments.

Will approach (3) work?  It depends on whether ixml’s facilities for
pragmas are up to the task.  How can we know whether a given design is
up to the task?  One plausible approach: try to sketch out what it
would look like to support namespaces, or text insertion, or attribute
grammars with synthetic attributes, using the proposed design.

So the pragmas proposal Tom and I developed has illustrations showing
that the pragmas proposal we are putting forward is capable of
handling the kinds of extensions to ixml that people have in fact
mentioned as potentially interesting or desirable.  The examples given
in the section “Worked examples” are just that: examples of how one
might use the capabilities of pragmas to do what one might want to do.

I do not know how even a cursory glance at the document could give
anyone the idea that the worked examples were proposals for additions
to the ixml spec.  Nor am I sympathetic to the principle that a
proposal for pragmas has no business showing examples of how pragmas
might be used.

-CMSMcQ

Received on Wednesday, 15 December 2021 19:59:40 UTC