Minutes IETF100: cdni
Content Delivery Networks Interconnection
||Minutes IETF100: cdni
CDNI Working Group Minutes
- Chaired by Kevin Ma (remote) and Phil Sorber (on behalf of Francois Le
Faucheur) - Meeting notes captured by Magnus Westerlund and Kent Leung, edited
by Kevin J. Ma - Audio Recording at:
https://www.ietf.org/audio/ietf100/ietf100-sophia-20171116-1810.mp3 Slides at:
Chair slides (Kevin J. Ma)
- No Document Updates
URI Signing (Phil Sorber)
Issue #28 - no comments; matt nodded
ISsue #30 - no comments
ISsue #31 - no comments
ISsue #32 - no comments
- Kent Leung: It is the simple glob? Rational for simple glob is gone. On the
fence if this should be removed or not. - Phil: There was no reason to remove
it; but this is possibly a reason to remove. Also on the fence. - Chris
Lemmons: Another issue is that the sub claim is designed for the "subject"
which seems to be interpreted as the user bearing the token. - Phil: Colon
issue is much more cut and dry - Matt Miller: Subject is ment for the entity
claims are over (URI) - Kevin Ma (as individual): in favor of CDNI-specific
sub-claims. Don't see a reason to keep the simple glob. Historically we added
simple glob because we didn't want regex. - Phil: Creating a pull request that
will remove the simple glob. Allows people to review that. - Chris: Subject
usage gleaned from usage, not from spec. - Kent: In favor of a pull request to
review, and remove if no objection.
- Kevin (as chair): will we be ready for last call in London
- Phil: Will publish after real usage in Feb. Optimistic about having a draft
ready for last call in London.
HTTPS delegation (Frederic Fieau)
Discussion of options on Slide 8:
- Robert Mugabe: Do you need to repeat the delegation metatdata for each pass?
you could make it a different resource/URI and link to it? - Frederic thinks
option 1 would easier to use. - Kevin: listing the paths doesn't add much;
could use links. making custom objects goes against the generic architecture
(with multiple metadata applied to paths rather than paths per metadata) the
metadata authors were trying to achieve. More information why this would be
necessary would help.
SVA OC Extensions (Ori Finkelman)
request routing - no comments
content management - scheduled triggers
- Phil: is the time not already UTC.
- Ori: The goal is to have things happening across multiple time zones at the
same time in the local time, like 3 am local time. Do not want geo-limited
triggers, per timezone with different UTCs. - Glen Deen: you want to send out a
generic time and have caches interpret it locally. Actually, in the context of
the client? What if a client is redirected from Boston to Denver? - Ori: This
is for bulk pre-positioning content in off peak hours. - Glen: This is useful
to have in CDNI. May need to consider a cache serving other timezones. Maybe
merge with geo-tags. - Sanjay Mishra: this is intended as a
restriction/instruction on the local cache serving the local environment.
uri signing - relayed token authentication
- Phil: this proposal is cleaver. But, there are things that could be
improved. Content specific metadata makes it less portable. Would like to
merge with URI signing and not require session state. Believe surrogate
validation is possible, with traditional URI signing. - Ori: Keep CP logic in
uCDN, but update dCDN token handling. - Phil: authenticate with uCDN; dCDN
generates its own token to client; TTL for re-authentication with uCDN.
logging - ran out of time
- Kevin: everyone should go an read the draft.
Based on today's discussion the new work being proposed is beyond the current
WG scope, and we would need rechartering.
Is the WG interested in taking on HTTPS delegation? Hmmm
- A weak (2?) hum in favor, a even weaker (1?) hum against.
Is the WG interested in working on SVA OC extensions? Hmmm
- Several hums in favor, and none against.