Re: defaulting

/ Jeni Tennison <jeni@jenitennison.com> was heard to say:
| Ah right, so you expect non-default <p:input>s that are unbound to
| default to an empty sequence. (And therefore generate a (static) error
| if they're not declared to accept a sequence.)
|
| I don't violently object to that; it just wasn't specified in the proposal.

I think I do object.

First, I don't want this

  <p:step>
    <p:input port="something"/>
  </p:step>

to be semantically different from

  <p:step/>

That is, I don't mind if authors put in empty inputs, but I don't want
it to be different from just leaving the input out.

Second, I don't want pipeline authors to have to know which ports are
default inputs and which are not. The current rule is very clear, any
unbound input port is bound to the default input port. Simple, clear,
understandable. Changing that rule to something like "Unbound default
input ports are bound to the default input port while unbound
non-default input ports are bound to the empty sequence" seems like a
real step backwards in usability.

I don't see anything about the defaulting story for pipelines that
requires this change.

I have a feeling the p:validate-xml-schema step may need to be
reworked, though. First, it probably needs a "don't use schema
location hints" option and second, it may need to have two forms, one
with a schema port and one without. Or maybe not.

| For the record, I think it would be worthwhile for the spec to note
| that steps declared with <p:declare-step> *don't* have any inputs or
| outputs auto-declared.

Yes, assuming we adopt this proposal, that's a good point.

                                        Be seeing you,
                                          norm

-- 
Norman Walsh <ndw@nwalsh.com> | When told of a man who had acquired
http://nwalsh.com/            | great wealth, a sage replied, 'Has he
                              | also acquired the days in which to
                              | spend it?'--Solomon Ibn Gabirol