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

Re: <input> for <pipeline>

From: Norman Walsh <ndw@nwalsh.com>
Date: Fri, 28 Sep 2007 11:42:57 -0400
To: public-xml-processing-model-wg@w3.org
Message-ID: <m2ir5u23um.fsf@nwalsh.com>
/ Richard Tobin <richard@inf.ed.ac.uk> was heard to say:
|> | After more thought, I am convinced that this just shouldn't be
|> | allowed.  A p:pipeline is effectively a step declaration along with
|> | implementation of the step.  It makes no more sense to allow binding
|> | of inputs on p:pipeline than on p:declare-step.
|> A binding on an p:input to a p:pipeline is intended to function as a
|> default binding if the processor doesn't provide one.
| Does it say that anywhere in the spec?

Uhm...no, apparently not.

|> Doesn't that seem like useful functionality?
| Possibly, but if so why does it only to pipelines?  Why not
| declare-step too?

I think only because you might have multiple uses of the same step
type, so the utility of defaulted inputs seems less likely.

| And how does it interact with connecting the input to the default
| readable port?

I don't think it does. The initial pipeline's bindings don't come from
a default readable port because there isn't one. If the impl. binds
stdin to one of the inputs, then stdin wins over the default.

If a pipeline is called as a step, I'd expect the default bindings to
occur and take precedence.

I wouldn't be surprised if that hasn't been clearly spec'd.

                                        Be seeing you,

Norman Walsh <ndw@nwalsh.com> | Worrying is the most natural and
http://nwalsh.com/            | spontaneous of all human functions. It
                              | is time to acknowledge this, perhaps
                              | even to learn to do it better.--Lewis
                              | Thomas

Received on Friday, 28 September 2007 15:43:16 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 21:32:44 UTC