Re: [CDNi] Dependencies & interactions between FCI and LI/MI/RI/CI

[Kevin Ma J <[e-mail]>]

Hi all,

  The previous question was how to divide up the work of advertisement of capabilities
  (i.e., what logging fields do you support, what metadata/options do you support, etc.).
  Ideally, capability advertisement would be defined with interfaces themselves, however,
  the FCI specifications are not as mature as other (MI/LI/RI/CI) interfaces.  In an effort
  to allow progress of those interfaces, as we finish defining FCI, we propose the following:

  1. The MI/LI/RI/CI drafts remove any place holders for FCI discussion.
  2. The FCI semantics draft will define the template for FCI objects.
  3. The FCI semantics draft will define a registry for FCI objects.
  4. We will publish separate RFCs describing capability advertisement for MI/LI/RI/CI.
     The FCI objects defined in these RFCs will be registered in the FCI object registry
     created by the FCI semantics draft.

  We believe this provides the best trade-off between expediency and functionality.

thanx.

--  Kevin J. Ma

> -----Original Message-----
> From: Francois Le Faucheur (flefauch) [mailto:[e-mail]]
> Sent: Friday, October 17, 2014 11:56 AM
> To: [e-mail]
> Cc: Francois Le Faucheur (flefauch); Jon Peterson; Jan Seedorf; Kevin Ma J;
> Daryl Malas
> Subject: Dependencies & interactions between FCI and LI/MI/RI/CI
> 
> Folks,
> 
> During the last IETF meeting we discussed dependencies and interactions
> between Fottprint & Capabilities advertisement and the other CDNI interfaces
> (Logging, Metadata, Redirection, Control): See excerpt from the Toronto CDNI
> minutes at the bottom of this message.
> 
> Two options for handling this were presented in Toronto and a tentative
> agreement for the second option was reached.
> 
> Since then, a bunch of the protagonists (Kevin, Jan, Jon, Daryl, Francois)
> continued discuss this and came up with a third Option which we feel is a
> better compromise. I have asked Kevin to summarise this third option and its
> merits for review and discussion on the list. If no objections are raised
> within a week of Kevin describing it, we'd like to adopt this as the
> approach going forward. Some of the LI/MI/RI/CI specs are on hold since they
> need to be aligned to the final agreed approach.
> 
> Please keep an eye for Kevin's message and raise support or concern
> promptly.
> 
> Thanks
> 
> Francois
> CDNI WG Chair
> 
> 
> 
> ========================
> FCI Semantics, draft-ietf-cdni-footprint-capabilities-semantics-03: Kevin Ma
> ----------------------------------------------------------------------------
> ---
> - Kevin presented the status update
> - Problem presented regarding, "Binary options for Logging/Metadata
> capability
> support is unlikely to be sufficient to describe optional modes." (See slide
> 3 for details)
> - Jan S: For things we know right now, binary is sufficient. But for future
> extensibility, we need to go beyond binary. Don't know?
> - Francois: Why not possible to avoid dependencies on FCI draft to publish
> Metadata and Logging drafts?
> - Kevin: Avoid shadow registries that are for FCI and other CDNI interfaces
> (metadata, logging, uri signing)
> - Jan S: Don't need to decide now, WG needs to be aware of the issue and
> post comments on ML. Hard to decide right now w/o thinking more about it.
> - Daryl: I would like to wrap this up ASAP and get Logging draft through
> - Francois: Is there a way to reference the advertisement objects in other
> drafts from FCI registry?
> - Kevin: This will block semantics draft, but not the FCI draft
> - Spencer: If you go with binary, and stuck in the future, what would you
> do?
> - Kevin/Jon: Rewrite the protocol, not non-backwards compatible
> - Spencer: I don't like that.
> - Jan/Kevin/Jon: More debate ...
> - Francois: Can you update FCI semantics draft so other drafts can document
> their own FCI advertisement objects? Give guidance to other drafts.
> - Matt C: My opinion that FCI is done before other drafts. I like option 2.
> - Cary Fitzgerald: Seems that you need to do the parsing?
> - Kevin: Duplicate registries are difficult to sync up
> - Kent: Option #2 seems reasonable
> - Ray: Likes option #2
> - Jan S: Timeframe issue, can FCI semantics be done in time?
> - Jon P: Need to steal stuff to write up FCI semantics
> - Jon P.: Next version of FCI semantics draft can be done by mid-Sept
> - Francois: Now the Logging draft needs to be updated to use the semantics
> for advertising of its objects. And there will still be a dependency for
> publication.
> - Room rough consensus on #2 to have each draft document its own
> advertisement objects using the FCI semantics
> - Chairs will identify a non-author reviewer for next rev of FCI semantics
> draft.