Re: Port references

Hi,

Norm Walsh wrote:
> / Jeni Tennison <jeni@jenitennison.com> was heard to say:
> | On the other hand, if there were a reason to name steps for other
> | purposes (e.g. to refer to them when indicating out-of-band
> | dependencies between steps) then I might change my mind again. :)
> 
> FWIW, if we needed to indicate out-of-band dependencies, I think I'd
> be inclined to do it by refrence to URIs produced/consumed rather than
> to the steps. But I might change my mind :-)

I'd like to see a way to do that as well, but I don't think it will 
cover all cases (e.g. when the out-of-band dependency is a set of 
documents with URIs that aren't known in advance).

Then again, given that we're giving no guarantees about whether steps 
might have side-effects or be functional, my current thinking is that we 
should simply say that the result of a pipeline must be as if the steps 
were conducted in sequence with each step fully completed before the 
next step starts, with pipeline engines free to optimise in any way they 
wish as long as they achieve that same result.

Effectively, this would punt responsibility for determining the 
interdependencies between steps to pipeline engines -- different engines 
will use different syntax to indicate out-of-band dependency between 
steps -- but I think that's probably OK for this version of the 
language. It would also mean that the ordering of the step definitions 
within the pipeline definition is significant, which is something to 
bear in mind.

Cheers,

Jeni
-- 
Jeni Tennison
http://www.jenitennison.com

Received on Wednesday, 10 May 2006 15:21:46 UTC