Sorry, I did not understand that last comment about "bush league"!
But yes, the spec is known to be flimsy and self-contradictory. In the end, we must make a choice on how to implement things in ARIA, while taking into account that other TOSCA implementations might interpret the spec differently. (Even more ideally: provide configuration options for ARIA's parser, so that it could better work with YAML files created for other TOSCA implementations. We have a few of these configuration options already.) This is exactly why the current PR for ARIA-1 is important: it introduces a broad test suite for TOSCA syntax and grammar, which obviously follows the interpretations we made for ARIA. But it can be run on other TOSCA parsers, too, at least giving us information as to where other parsers differ in their interpretations of the spec. I will say that our rules of thumb has generally been: 1) if a strict and loose interpretation are possible, choose the stricter one, and 2) keep the "spirit of the spec" in mind: object-orientation and enforcement of the parent type contract. On Tue, Nov 28, 2017 at 12:44 PM, DeWayne Filppi <dewa...@cloudify.co> wrote: > Wow. That is reeeeeeeeeeeaaaaaaaaaaaaaaaaaaaaaaally bad and bush league. > > On Tue, Nov 28, 2017 at 10:42 AM, Tal Liron <t...@cloudify.co> wrote: > > > "Examples" in the spec routinely break the syntax of the spec... I think > > it's best not to trust them. > > > > On Tue, Nov 28, 2017 at 12:40 PM, DeWayne Filppi <dewa...@cloudify.co> > > wrote: > > > > > I suppose it lets you name interfaces whatever you want, which is > > confusing > > > because of other areas of the spec. Note that there are tons of > examples > > > in the spec without the "type" specified. > > > > > > On Tue, Nov 28, 2017 at 10:27 AM, Tal Liron <t...@cloudify.co> wrote: > > > > > > > I mentioned this to you in the previous thread: the "type" field is > > > > required for interface definitions according to TOSCA syntax. So, > even > > if > > > > it's the same as what you are inheriting, you must specify it. > > > > > > > > On Tue, Nov 28, 2017 at 12:04 PM, DeWayne Filppi < > dewa...@cloudify.co> > > > > wrote: > > > > > > > > > Now that the 'subclassing' problem has been resolved, overriding > > > > interface > > > > > methods is breaking. Simple example: > > > > > > > > > > tosca_definitions_version: tosca_simple_yaml_1_0 > > > > > > > > > > imports: > > > > > > > > > > - aria-1.0 > > > > > > > > > > node_types: > > > > > > > > > > T1: > > > > > derived_from: tosca.nodes.Root > > > > > interfaces: > > > > > Standard: > > > > > create: > > > > > implementation: > > > > > primary: i1.sh > > > > > delete: > > > > > implementation: > > > > > primary: i1.sh > > > > > > > > > > The error, using Aria in the ARIA-1 branch: > > > > > > > > > > Validation issues: > > > > > 2: required field "type" in > > > > > "aria_extension_tosca.simple_v1_0.definitions.InterfaceDefinition" > > > does > > > > > not > > > > > have a value > > > > > > > > > > > > > > >
