Re: Syntax

/ Alessandro Vernet <avernet@orbeon.com> was heard to say:
| On 5/17/06, Norman Walsh <Norman.Walsh@sun.com> wrote:
|> We've been kicking syntax around for a bit. Unfortunately, I don't see
|> consensus forming anywhere in particular. I think I still prefer a
|> simple, syntactically regular formulation for "ordinary" components:
|>
|> <p:step name="QName">
|>   <p:param name="QName" .../>
|>   <p:input name="QName" href="someURI"/>
|>   <p:input name="QName" ref="someLabel"/>
|>   <p:output name="QName" label="aLabel"/>
|> </p:step>
|>
|> I'm sticking with the "user names each output label" for the time
|> being. There are a number of shortcuts we could introduce:
|>
|> 1. Multiple references to the same label implies a "tee"
|> 2. href on a p:output implies a "save"
|
| I agree on everything above.
|
|> 3. Omitting exactly one required p:output on a step and exactly one
|>    required p:input on the following step implies that that output
|>    is connected to that input. (Alternatively, components could nominate
|>    default inputs and outputs and those could be connected together by
|>    default.)
|
| Not sure about this though. We went back and forth on the issue of
| implicit connections between steps.

FWIW, I'm not sure about any of these shortcuts. I wasn't proposing
them with any conviction, just enumerating the ones that I recall we
discussed.

|>   <p:peephole>
|>     <p:step/>
|>   </p:peephole>
|
| What would be the semantic of the peephole?

Peephole (we really need a better name) is the component that takes an
XPath expression that (presumably) identifies an element and applies
the steps to that subtree.

You could, for example, apply XSLT to each of the <chapter> elements
in a book without ever having to load the entire book into memory
(assuming your implementation was smart enough to evaluate the
expression in a streaming fashion).

                                        Be seeing you,
                                          norm

-- 
Norman Walsh
XML Standards Architect
Sun Microsystems, Inc.

Received on Friday, 19 May 2006 19:20:33 UTC