W3C home > Mailing lists > Public > public-xml-processing-model-wg@w3.org > June 2007

defaulting

From: Richard Tobin <richard@inf.ed.ac.uk>
Date: Fri, 22 Jun 2007 16:01:23 +0100 (BST)
To: public-xml-processing-model-wg@w3.org
Message-Id: <20070622150123.72C32225642@macpro.inf.ed.ac.uk>

Further to discussion during yesterday's telcon, herewith a proposal
about input and output defaulting for pipelines.

It basically adopts Richard's suggestion that a pipeline's inputs/outputs
default if there is a need for them to do so, because their first
component needs an input/final component produces an output.

Here's how this works in detail:

1) We make defaulting more symmetrical for p:input and p:output, by
   adding a 'default' attribute to p:input, allowed only in
   declaration contexts (as there already is one on p:output).  An
   input to a step is _the_ default input iff

  a) it is the only declared input and it was not declared with
     default='no';
or
  b) it is declared with default='yes'.

  It is an error to declare more than one input as default='yes'.
  It is _not_ an error to have no default inputs.
  It is _not_ an error to have no default outputs (this is a change).

  The default input of a step is bound to the default readable port if
  it is not otherwise bound;

  Non-default inputs are only bound to the default readable port if
  you call for that to happen by writing
   <p:input port="not-the-default-input-port"/>

2) Pipelines w/o a declared default input get an un-named default
   input iff their first contained step has an unbound default input.

3) Pipelines w/o a declared default output get an un-named default
   output iff their last contained step has an unbound default output.

This means that not only does

 <p:pipeline>
  <p:identity/>
 </p:pipeline>

do the obviously right thing, but so does

 <p:pipeline name='top'>
  <p:input port='stylesheet' default='no'/>
  <p:xslt>
   <p:input port='stylesheet'>
     <p:pipe step='top' port='stylesheet'/>
   </p:input>
  </p:xslt>
 </p:pipeline>

To construct a pipeline which has no input, just start it with a step
which needs no default input, e.g.

 <p:pipeline>
  <p:identity>
   <p:input port='source'>
    <p:document href="http://www.example.org/fixedInput.xml"/>
   </p:input>
  </p:identity>
  .  . .
 </p:pipeline>

This is the obvious case for a pipeline with no input, that is, a
pipeline with a fixed input.

Exactly parallel, to construct a pipeline which produces no output,
just end it with a step which produces no default output.

We currently don't have an easy way to do this, but by analogy with
the previous (pipeline with fixed input) example, a pipeline with
fixed output _ought_ to be such a case:

 <p:pipeline>
  . . .
  <p:store href="http://www.example.org/fixedOutput.xml"/>
 </p:pipeline>

To accomplish this, we propose two additional related changes:

4) The declaration for p:store should change to look like this:

 <p:declare-step type="p:store">
  <p:input port="source"/>
  <p:output port="result" default="no"/>
  . . .
 </p:declare-step>

 That is, p:store _has_ no default output.

5) It _is_ an error for the _default_ output of a step to fall on the
floor.

We believe that making these changes actually cleans up a number of
minor glitches, and overall makes the spec. 

-- Richard and Henry
Received on Friday, 22 June 2007 15:01:35 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 8 January 2008 14:21:53 GMT