Minutes IETF112: cdni
minutes-112-cdni-02

CDNI WG Minutes
IETF-1112 Online
Chairs: Kevin Ma and Sanjay Mishra
AD: Francesca Palombini

Agenda and Slides: https://datatracker.ietf.org/meeting/112/session/cdni
Recording: http://www.meetecho.com/ietf112/recordings#CDNI

Chair slides (Kevin J. Ma)
--------------------------

Chris Lemmons: volunteered to jabber scribe

Kevin: last call for adoption on
draft-sopher-cdni-triggersextensions-rfc8007bis: no objections

Kevin: call to add a milestone: no objections

Nir: agreed to a target deadline of March 2022

Kevin: last call for adoption on draft-sopher-cdnifootprint-types-extensions:
no objections

Kevin: call to add a milestone: no objections

Nir: agreed to a target deadline of March 2022

CDNI Footprints (Nir Sopher)
----------------------------

Nir: only update since July was to rename ISO3166-2Code to SubdivisionCode

CDNI Triggers (Nir Sopher)
--------------------------

Nir: no updates since July; wanted to complete adoption before proposing
additional changes

Nir: new proposal as discussed on the list: replace individual trigger
properties with an array of generic trigger objects (a la generic metadata
objects)

Kevin: agreed that an array of generic objects makes more sense for
extensibility

Kevin: agreed with pushing the update to a working group draft and then pushing
a new revision with the proposed changes to the trigger object

CDNI URI Signing (Phil Sorber)
------------------------------

Phil: requested help on addition guidance for designated experts

Kevin: offered to take a look

Phil: three open questions to the WG (to be sent to the list as well after the
meeting)

Question 1: should we remove client IP?

Andrew Ryan: noted valid use cases where client IP is useful; would rather keep
it.

Kevin: noted there are valid internal network use cases were the reason it was
added. the known limitations of using client IP do not negate those use cases.
Could add more disclaimers, would rather keep it.

Chris: agreed client IP has valid use cases, but also noted that an array of
IPs might be more useful (along with the ability to combine IPv4 and IPv6), but
the change may not be warranted at this late stage.

Question 2: should we remove shared key?

Kevin: noted there are valid internal use cases as well as legacy compatibility
reasons for supporting it.

Chris: could just remove it.  internal use cases do not need interop; and it is
obvious how to do it if someone wanted to use shared keys.

Question 3: should we make URI Container mandatory?

Chris: noted a valid use case for multiple issuers.  requiring inclusion of a
wildcard URI Container is a waste of CPU on a regex that could otherwise be
omitted.

Chris: posted the PR for proposed text on path-style and form-style params

Phil: wanted to get more WG feedback on Chris's text before merging it

Kevin: to review

Glenn Goldstein: asked about common access token CTA-WAVE/SVA compatibility

Kevin/Phil: not familiar with the spec; will need to review

Kevin: URI signing is not going to be compatible with all legacy algorithms,
e.g., AWS signature v4

Kevin: URI signing was added to the December telechat agenda; when do we need
to resolve questions by?

Francesca Palombini: need to post updated draft by 11/25 to give ADs time to
review; otherwise we can push it out to the next telechat.

HTTPS Delegation (Frederic Fieau)
---------------------------------

Kevin: posted comments on the new draft on the list; need to address before
discussing last call

Sanjay: agreed not ready for last call.  the RFC9115 change could mean a lot of
changes for how the dCDN handles delegation.

Kevin: lets wait for and review the next version of the draft and then
re-evaluate draft readiness

Capacity Advertisement: (Andrew Ryan)
-------------------------------------

Andrew: SVA wants to reuse the FCI interface for capacity advertisement

Andrew: the long term goal is to define a near real time telemetry interface,
but the short term goal is to describe generic telemetry sources.

Andrew: there are existing telemetry interfaces in dCDNs that uCDNs need to
access.

Kevin: are the existing sources proprietary or using standard protocols?

Andrew: it's mostly HTTP polling of bespoke interfaces with bespoke formats

Kevin: the metrics are common even though interfaces are not yet standard?

Andrew: the metrics are generally well understood.  can create a registry for
the metrics.  but how they are calculated introduces ambiguity, so both the
uCDN and dCDN must use the dCDN metrics

Andrew: the FCI object allows advertising telemetry for an entire CDN or a
specific host.

Andrew: the MI only allows scoping to a host and not all hosts.

Kevin: MI is not idea for this use case, but you tried to fit it into the
existing MI?

Andrew: agreed.  the proposal is a middle ground to adhere to the available
framework, but would welcome feedback or suggestions to work around it.

CDNI Metadata (Glenn Goldstein)
-------------------------------

Glenn: SVA wants to extend the MI to handle more use cases

Glenn: All changes are implemented in new generic metadata objects

Glenn: one issue is with service id.  service id as a host metadata is
backward.  usually a CDN has a service id with multiple hosts under it.  the MI
does not support grouping hosts.

Glenn/Alfonso: re: Kevin's comments on the list, origin caching rules are often
different from CDN caching rules and the uCDN acts as an origin to the dCDN, so
it is easier for the uCDN to specify different rules to the dCDN, rather than
respond differently to a dCDN.

Glenn: processing stages: processing can happened in four places (receiving a
request from a client, forwarding a request to a uCDN, receiving a response
from a uCDN, and serving a response to the client from cache).

Glenn: processing may include transforming a request, transforming a response,
or generating a synthetic response

Glenn: were able to fit all of the processing stage configuration as generic
metadata objects

Kevin: impressive that it all fit into metadata.  is this the right way, or
just how it had to be to fit into the MI?

Glenn: it fits and it's clean

Glenn: SVA is extending the MI to support publish and delete

Kevin: MI was originally designed for triggered invalidation.  is SVA going to
bring back the publish and delete proposals to the WG?

Glenn: will discuss with Sanjay and the SVA

Alfonso: the extensions are probably compatible with triggers.  can start a
discussion on the list.

Glenn: additional extensions for lifecycle managment: versions, staging vs prod
deployments, rollbacks, etc.

Kevin: the draft is big.  perhaps break it up into separate drafts grouping
similar metadata to make it easier to digest with less context switching? 
could speed up adoption of more straightforward parts.

Glenn: agreed with grouping relevant metadata together

Sanjay: agreed with grouping relevant metadata together

Session closed.