RE: Single implementation, multiple annotations (model testing)

Section 5 of the data model does accommodate a Annotation Collection with an Embedded Page (which can contain annotations), so in theory there is a structure that could ship multiple annotations. But the Collection+Page approach will not be implemented by all annotation clients and is itself optional – i.e., you can have an annotation Collection that contains no annotations, simply points at the first page of annotations in the collection which then has the actual annotations as part of an items array. 
In any event the tests and assertions covering sections 1-4 of the model as written are not designed to test / validate multiple annotations at once.  These assertions could of course be referenced by schemas that are designed to test / validate pages of annotations. This would not be all that difficult, but as mentioned above would require implementers not implementing section 5 of the model to manually wrap their annotations within annotation page structure, which itself has 8 potential keys. Seems like it would greatly reduce those interested in implementation testing and introduce opportunities for errors which would obscure actual counts of implementations of model features from sections 1-4.
As configured now, if you paste multiple annotations concatenated together into the text box for a test, the tests basically refuse to run.
If you wrap the multiple annotations within some array (e.g., "allOf"), the tests run but you fail all of the implementation check tests and only many of the validation test assertions – you pass some of the value validation assertions (based on the check not finding the key associated with the value validation rule, and therefore not having anything to check the rule against).
So I think we have to live with multiple submittals from a single implementation when that implementation allows for creation of different kinds of annotations each kind of which implements different parts of the model.
Tim Cole
From: Shane McCarron [] 
Sent: Thursday, September 08, 2016 9:21 AM
To: Cole, Timothy W <>
Cc: W3C Public Annotation List <>
Subject: Re: Single implementation, multiple annotations (model testing)
It makes sense, and I think that is fine as long as we add entries to the index to explain what each subcode means.
Alternately, I would be interested in seeing what happens if multiple annotations, each with a variation as you mentioned, are included in a single message.  I think that the data model accomodates this.  If it were done that way, would the test just "pass" for all of the various features ?
On Thu, Sep 8, 2016 at 8:21 AM, Cole, Timothy W < <> > wrote:
It is commonplace for a single annotation client to generate multiple kinds of annotations. Each kind of annotation from a given implementation may implement different features of the model. So to capture this information, an implementer will need to run multiple annotations through our test suite. As best I can tell this means multiple test results reports, each of which will need to be submitted to <>  each resulting in a separate column in the all.html report.
First, is my understanding corrrect, or is there some way to combine test result reports?
If my understanding is correct, is it therefore appropriate to use the same 2 letter prefix (representing implementation), and then label each report with its own 2-digit number? So for Janina's emblem annotation client which implements 3 kinds of annotations, she would have EB01, EB02, EB03.  EB01 might implement a embedded textual body, while EB02 might implement an external Web resource as body, and so on.
Does this make sense?
Tim Cole

Shane McCarron
Projects Manager, Spec-Ops

Received on Thursday, 8 September 2016 16:03:50 UTC