Re: Feedback from DocEng 2006

Richard Tobin wrote:
>> 2. Why no loops/recursion?
> 
> Pipelines can be recursive in the sense that there will (I assume) be
> a "call pipeline" component that could call the same pipeline.
> 
> Of course there is looping too, over document sequences and nodes, but
> presumably they want looping with the output feeding back to the input.

One use case that we talked about awhile ago and decided not to put in
our V1 requirements and use case document was the example of following
the "next" relation on an Atom feed.   This would require the ability to
test for the existence of that element and continue to loop while it
is found in the document.

In addition, such loops might have complicated processing semantics in
terms of what results are produced from the loop.  In the case of
the Atom "next" use case, you really want one feed composed of all
the "entry" elements in the order given to you by the server as if
you retrieved each feed in order.  That means you have to expand the
"link" element you received at the start of the document at the very
end of the document.  That's easy to write in a regular programming
language but hard (or at least harder) to describe in a pipeline.

While it is a neat idea and certainly useful, I can't say that it is a
V1 requirement that we do this.  Which is what we decided already.

I'd be interested in hearing about other loops that we've
decided--explicitly or implicitly--that we're not doing in V1.

--Alex Milowski

Received on Tuesday, 24 October 2006 02:49:52 UTC