W3C home > Mailing lists > Public > public-xml-processing-model-wg@w3.org > November 2008

Re: CR draft ready for review

From: Norman Walsh <ndw@nwalsh.com>
Date: Fri, 21 Nov 2008 08:33:47 -0500
To: public-xml-processing-model-wg@w3.org
Message-ID: <m2k5axp5j8.fsf@nwalsh.com>
Toman_Vojtech@emc.com writes:
> Ooops, I missed this somehow. I guess I have to fix this in our
> implementation. Thanks for a bug report. :)

:-)

> One question, though, just to be sure: If I use p:declare-step instead
> of p:pipeline, and I declare a primary parameter input port there, we
> don't manufacture the default binding to it in that case?

We manufacture default bindings to steps that expect parameters. This
pipeline works just like you'd expect if you wrote it with p:pipeline:

<p:declare-step name="main" ...>
  <p:input port="source"/>
  <p:input port="params" kind="parameter"/>
  <p:output port="result"/>

  <p:xslt>
    <p:input port="stylesheet">...</p:input>
  </p:xslt>
</p:declare-step/>

What we don't manufacture is the p:sink step if you've explicitly
created a parameter input port and then *never* actually used it.

> Ah, so it is about something connecting to the parameter input port, not
> that the port must be connected to something. But I thought that inside
> a compound step, you don't have to connect anything to the declared
> input steps. Only when you *use* the step, you have to connect its input
> ports to something... I thought that pipelines like the following would
> run just fine (the compound step just ignores any data/parameters it
> gets):
>
> <p:declare-step ...>
>   <p:input port="source" primary="true"/>
>   <p:input port="another-source"/>
>   <p:input port="params" kind="parameter"/>
>   <p:output port="result"/>
>   <p:identity>
>     <p:input port="source"><p:document href="..."/></p:inline>
>   </p:identity>
> </p:declare-step/>
>
> Sorry, I am probaly just under attack of my recurring confusion about
> what "connected" and "binding" actually means in XProc...

To my surprise, I think you're right (I'm not surprised you were
right, I'm surprised about the spec :-). No where in our spec does it
say that *input* ports have to be bound. I wonder if it should? I
think probably not. I wonder if it used to, or if I just imagined that
it did? In any event, unless we change the spec to say that (primary)
input ports must be bound, I think we can drop that paragraph about
automatic binding to p:sink for parameter input ports on p:pipeline
that aren't used.

Good catch, Vojtech!

                                        Be seeing you,
                                          norm

-- 
Norman Walsh <ndw@nwalsh.com> | If brute force doesn't work, maybe
http://nwalsh.com/            | you're not using enough brute force.

Received on Friday, 21 November 2008 13:34:30 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Friday, 21 November 2008 13:34:30 GMT