Re: manifest based processing

Le 19/02/2014 17:44, Florent Georges a écrit :
> On 19 February 2014 17:33, Olivier JEULIN wrote:
>
>> If I understood correctly,  you want to automatically connect an
>> output named "A" to an input named "A"?
>   Nop, not based on the name, that looks too fragile to me (plus, you
> want to connect an output port "result" to an input port "source").
> That is the raison d'etre for the kinds.  They create classes of
> "stuff flowing through ports", and we define that two ports will by
> default connect all their ports of the same kind (the user still has
> the choice to define the connections manually).

My point exactly: "stuff flowing through ports" = you attach a name to
a pipeline (like "hot"/"cold" in the bathroom, or "phase"/"ground" for
electrical cables).
So there is no point to have another naming for source/result.
result(kind-X) |=>=| source(kind-X)
is equivalent to:
kind-X  |=>=| kind-X
Lets cut the middle man 
Of course, we still need to be able to make an explicit connection when
needed.

This is similar to pipes in unix: programs have stdin/stout and they
are connected by default.


-- 
Olivier Jeulin

Received on Wednesday, 19 February 2014 20:47:36 UTC