[dsdl-comment] Re: FAQ: (1) NVDL versus XProc, (2) Performance of NVDL

Costello, Roger L. wrote:
> Hi Folks,
>
> 1. I just saw this on a web site:
>
> "I'm personally against NVDL, I think I would rather have XProc used
> to deliver multiple schema languages because then I also have all the
> benefits of XProc in my architecture."
>
> I don't know enough about XProc to respond. Can someone who is
> familiar with XProc respond to this?
XProc and NVDL address different layers: it is not that one can
susbstitute for the other but rather than one can implement the other
given: in particular, an NVDL schema could (I expect) be transformed
into an XProc running XSLTs for extracting fragments then dispatching
them to various validators.

So that comment is just like
"I personally wouldn't use Schematron, I would much rather have XSLT to
validate because then I also have all the benefits of XSLT in my
validations"

Michael Kay made a comment like this a while ago, so it is not an
unrespectable opinion, and there is no reason to object that anyone
*must* use Schematron (or NVDL), nor that it will be optimal in all
circumstances, nor that some validations might require more. Indeed,
this is one reason ISO DSDL has a slot reserved for an execution
framework: we are waiting on the XProc results to evaluate it, and the
XProc WG kindly took on board some SC34 DSDL requirements.

However, the more that constraints are expressed at a high level, the
more they can be re-purposed. This is XML 101 of course. For example,
think how difficult it is to repurpose the XPaths in Schematron which
are so low level, while grammars allow all kinds of generating thing:
data binding, GUIs, diagrams.

> 2. Another comment that I heard is:
>
> "The NVDL processor (dispatcher) adds an additional layer and thus
> introduces a performance penalty.
>
> Does anyone have performance measurements that refutes (or
> supports) this statement?
Penalty compared to what? Without a base case, it is meaningless.
Layering does not intrinsically add any performance penalty, since
layered declarations may be implemented folded into the code. What
layering does is allow more optimizations than general-purpose
expression: this was the basis of 4GLs. Think SQL versus Basic.

Cheers
Rick Jelliffe

--
DSDL comments
To unsubscribe, please send a message with the
command  "unsubscribe" to dsdl-comment-request@dsdl.org
(mailto:dsdl-comment-request@dsdl.org?Subject=unsubscribe)