Never mind, I already got an answer from the spec =)

On Wed, Nov 22, 2017 at 6:29 PM, Tal Liron <[email protected]> wrote:

> Less cumbersome, sure. But it also breaks the object-oriented contract.
>
> Imagine this situation: a third party develops a powerful node type (let's
> say: a virtual router) with many well-defined and polished operations on
> custom interfaces (with their own types), custom workflows for various
> advanced features, and even comes with custom ARIA-based tooling for
> real-time analysis and monitoring.
>
> Someone can come along and create a node template based on this node type
> that changes the operation inputs. For this to work properly, that someone
> *should* also provide a new operation implementation that would deal with
> these different inputs. But, these inputs must be sent for this to work: so
> the custom workflows must also be changed, as well as the custom tooling.
> And if new tools come along from that that third party, they will not work.
>
> Object-oriented is cumbersome, sure. :) But that's also the design
> principle of TOSCA, and this is a case that breaks it. We either want our
> node types to have extensible contracts, or allow them to have breakable
> contracts. It doesn't make sense to me to be somewhere in between:
> strictness should be all the way.
>
> Finally, I don't understand this question:
>
> On Wed, Nov 22, 2017 at 7:21 AM, Avia Efrat <[email protected]> wrote:
>
> > By the way, do you see a way of not being able to derive the value of an
> > intrinsic function's 'result'? (I know it is not currently supported)
> >
>

Reply via email to