Yes, this is more or less the P4Info I would expect. But my P4 is a bit rusty...

• IngressI is used to prefix all objects defined in the IngressI control block. The way I see it is that when you instantiate the V1Switch package, you instantiate an anonymous instance of IngressI(), while @name("altIngressName") is an annotation on the type. I don't remember if the following syntax is valid, but this would be my first instinct as to the "right" syntax if you want to introduce a new name for the instance:

V1Switch(ParserI(), VerifyChecksumI(), @name("newName") IngressI(), EgressI(),
         ComputeChecksumI(), DeparserI()) main;

Now one could argue that we should still use the annotation on the control block if that's all we have, and I would be fine with that. Another valid option would be to use the parameter name, which is ig in this case (actually this is what I would have guessed we were doing before looking at your example).

• The annotations on the action references in the table declaration are ignored. I would say this is a limitation of the ActionRef message in P4Info which should probably include a name field. The logic for having a dedicated ActionRef message was to be able to accommodate for something like the code below (again not sure if this is currently legal P4). I think this is a change we should consider making in P4Info.

table t {
    actions = {
       @name("foo") a(5);
       @name("bar") a(9);
    }
}

Besides these 2 points, I don't think anything else is contentious. The @name annotations are handled as they should for match fields, action parameters, actions, tables, etc.

I feel like an update to https://github.com/p4lang/p4runtime/blob/main/docs/v1/guidance-for-generating-p4info.md is coming :)