Re: [tosca] intrinsic functions inside metadata?

Well, at least until TOSCA 1.3 this was not in spec, so I think this could at best be a proposal for TOSCA 2.0.

If we do so, I'm not entirely sure this would be a good idea. One of the things we discussed (but don't think we decided on) was that metadata values could be any arbitrary unvalided YAML, not just strings. If we decide that it's actually a value ... well, it's untyped. We discussed maybe supporting an "any" type in TOSCA, so that might be what happens here.

But also, I'm concerned that this could lead to poor designs, a way to "sideload" properties into node templates and breaking the OOP contract. What's next, a get_metadata function? Once we allow metadata to parameterized it becomes ... data. This is not necessarily a terrible thing, but it's a big change and deserves more examination.

On Wed, Apr 21, 2021 at 11:57 AM Chris Lauwers <> wrote:

I received an interesting question about whether it is legal to use intrinsic functions inside metadata, as shown in the following example:

Â

Â

We have a node which looks as follows:

Â

ÂÂÂ c1_nic1:

ÂÂÂÂÂ type: cisco.nodes.nfv.VduCp

ÂÂÂÂÂ properties:

ÂÂÂÂÂÂÂ layer_protocols: [ ipv4 ]

ÂÂÂÂÂÂÂ order: 1

ÂÂÂÂÂÂÂ virtual_network_interface_requirements:

ÂÂÂÂÂÂÂÂÂ - support_mandatory: true

ÂÂÂÂÂ metadata:

ÂÂÂÂÂÂÂ security_groups: { get_input: VIM_NETWORK_MGMT_SEC_GRP_0 }

ÂÂÂÂÂ requirements:

ÂÂÂÂÂÂÂ - virtual_binding: c1

Â

But are being queried about the
security_groups value, with the concern being that âtosca functions are not allowed inside metadata.â

Â

I personally donât see any issue with this, but Iâm curious if anybody feels differently?

Â

Â

Thanks,

Â

Chris

Â

Â