Re: Steps with non-connected outputs

On 4/12/07, Alessandro Vernet <avernet@orbeon.com> wrote:
>
>
> On 4/12/07, Alex Milowski <alex@milowski.org> wrote:
> > > 1) Is this a valid pipeline?
> >
> > Yes.
>
> OK.
>
> > > 2) If it is, should the query be sent to the database (yes, no,
> > > implementation dependent)?
> >
> > Yes.
>
> Why would the step run in this case? Do we have something in our
> current specification that mandates this behavior?


All steps are started and ended in my model.  I don't think we can
get away from that and so if we're missing that, we need to fix that
in our specification.

That is, the pipeline starts and stops.  That bounds computation for
all steps.  If there is a step with no inputs and no outputs, it can still
"know" when the pipeline starts and stops.

I've been thinking we need to describe the step's view of the time
sequence of processing.  This clarifies for me that we need to do that.

In my model, we have the following time sequence for the lifetime of
a step:

1. Initialization of the component happens.  This is typically while the
    pipeline itself is being initialized.  This is a event that probably
happens
    outside our specification.

2. Resources that are statically known and bound to input ports are made
    available to  steps.  This allows steps to detect errors staticially (
e.g. I can't
    locate or compile the XSLT transform).  This is an optimization but a
very
    important one.

3. Output ports are bound to their recipients.

4. The pipeline starts and the step is notified.

5. Some sequence of documents are received (or "pulled") on the
    input ports for the step.

6. The pipeline ends and the step is notified.


I think our specification needs to allow for 1-3 but needs to focus on
4-6 as standard semantics.


-- 
--Alex Milowski
"The excellence of grammar as a guide is proportional to the paucity of the
inflexions, i.e. to the degree of analysis effected by the language
considered."

Bertrand Russell in a footnote of Principles of Mathematics