Types of PWP manifests?

I'm slightly confused by the discussion we had during our call yesterday [1] about the possibility to define priorities in the manifest-retrieval algorithm.

My understanding of Ivan's description –please correct me if I'm wrong– is that in some way we define "abstract" manifests depending on how the manifest can be retrieved.
For instance, one of these "abstract" manifest is the manifest retrieved as a payload of the GET request. Regardless of what is the ultimate format and syntax, regardless of whether conneg is involved or not, the information is the same.
We can then prioritize which manifest has precedence when merging the various abstract manifests.

We also appear to have 5 types of abstract manifests:

- manifest linked from the HTML
- manifest embedded in the HTML
- manifest corresponding to the request payload
- manifest retrieved from the package
- manifest linked from HTTP link header

Am I correctly interpreting the current state of the discussion?

When I look at the diagram visualisation [2], the things that confuses me are:

- how do we define that the canonical locator resolves to a package or an HTML or a manifest?
- what's the "standalone" manifest compared to those linked from the HTML. They're both referenced as "M2" in the diagram, does it mean that they are equivalent?
- am I right to assume that the manifest retrieved from the package can also be a merge between a linked manifest and an embedded manifest (as with the M2=M1.0+M1,1 process described for HTML)?

Sorry if I muddy the water here, but I'm failing to grasp the big picture.

Romain.


[1] https://www.w3.org/2016/03/09-dpub-loc-minutes.html#item04
[2] https://raw.githubusercontent.com/w3c/dpub-pwp-loc/366ce8214e23389870251b58f204b4e591b00a1f/drafts/PWPClient3.png

Received on Thursday, 10 March 2016 17:53:44 UTC