Re: [oiic-formation-discuss] What is interoperability anyway / really?

["Dave Pawson" <dave.pawson@gmail.com>]

2008/6/8 David RR Webber (XML) <david@drrw.info>:
> Dave,
>
> Guess we need to broach this topic some too.
>
> For me interoperability is about people being able to predictably exchange
> documents - and publish clear instructions on rules / handling / content
> models including contextual templates and patterns that are intended /
> expected.

End user expectations? So again we need to care about how far back
into implementation we go. How might content models impact user
interaction and expectations David?



>
> Sadly we have a chunk of folks out there that believe publishing an XSD and
> then enforcing validation there of is how you achieve interoperability.

<grin/> I'd prefer Robs C++ as an example! XSD is quite possibly the worst
example of interop expectations!


>
> They are angry and confused when things pass their XSD and proceed to trash
> their backend applications - or frustrated when people construct what appear
> to be reasonable content from human perspective - but the XSD rejects as
> invalid for some arcane reason.  Then of course the final straw is wierd
> content created automatically by tooling software that passes the XSD but is
> a nightmare to process and contains 3x the amount of markup that sane people
> would ever want in there!

Good examples of bad interoperability! And why it's necessary.



>
> The problem put simply is that XSD is non-deterministic - and was never
> meant to be deterministic.
[[And it should be put down]]

>
> So in our case - scripting tools like CAM add context - and ability to
> assert deterministic rules that make predictable outcomes possible.  A
> "filter" in effect - that can tell you if the content will be acceptable and
> make sense.

Being totally ignorant of CAM, can you provide an example of
how such a tool would help us here please.
How it might help an end user have confidence that Vendor A
product produces the same  result as Vendor B?
  Even if it's not visual, an example would help us understand.

>
> But CAM is designed for business transaction exchanges - not documents.
>
> However the principles are the same - just that the set of "smart" functions
> need to be adapted to document application needs - such as page counts and
> section content / page headers / footers, signatures, et al - which are not
> part of the lexicon of CAM.

Understood, which is why I'm asking how your experience might be
brought to bear for this use case.


>
> So - overall seems to me for interoperability - you need the same ability as
> CAM brings to produce open standard sharable rule sets and patterns that can
> be verified, documented and will check instances conform to them - but with
> a document lexicon of functions....


Interop really is hard to pin down with easy clear words.
As for testing the test kit, meta testing!

This isn't easy!

regards


-- 
Dave Pawson
XSLT XSL-FO FAQ.
http://www.dpawson.co.uk