Delay-Tolerant Networking (dtn) WG

IETF 117, San Francisco
2023-07-26, 09:30-11:30 America, Pacific Time

Area Director: Zaheduzzaman Sarker (Erik Kline covering)
Chairs: Edward Birrane, Rick Taylor
Secretary: Adam Wiethuechter

Introduction, Charter/Milestones, WG Last Calls (5 mins)

Speaker: WG Chairs
Document: N/A

cRT: welcome! This is the "session title" WG.

Operating a queue for partipcation. Please use it!

cEB: review of agenda, no calls for agenda bashing.

Review of milstones

cEB: lots of items that need some energy

WGLC

cEB: in LC; COSE, Admin Records

Expected by end of session: IPN URI, DTNMA

Update of IPN URI Scheme (10 mins)

Speaker: Rick Taylor (Chair Hat Off)
Document: https://datatracker.ietf.org/doc/draft-ietf-dtn-ipn-update/

Run through 116 to now changes.

Registry will talk now.

Change of service definition. Not a generic type or protocol id

"well known": reasonable default assumed, public specified and widely
deployed

Scott B: direction is right, concern for interop.

Agree with interop issue and this is loose generalization. Analogous to
port numbers.

Scott B: you cant force something agree. suggest desire these provision
to have same force and intent as SHALL.

Alberto Montilla: do we need this specified now? risk of over
specification that is required and creating interop issues.

Agree with sentiment but providing facility for when it needs to be
there.

CBHE Service Numbers reg will exist and is not to be touched.

New reg is explicit for BPv7. Allocation policy to share even betweeen
private use and controlled values and levels of restriction. This is due
heavily to CBOR encoding size.

0-octet encoding is private use. 1-octet split (lower private use, upper
standards action), 2-octet (lower private use, upper specification
required (any SDO))

Marc Blanchet: like early days, we don't know what happen. we nay force
a stndard action and never gets deployed. make sure it is well managed.

Pragmatic approach is what we need.

Alberto: curious about intention that BPv7 would start as existing v6
reg?

No

Alberto: why?

Based on service definition carry across might not make sense. Also
encoding characteristics being different make this hard.

Alberto: space agencies will be "pseudo-standardizing"

This is not a breaking change! last was for BPv6 and everyone assumes
for BPv7. understand pain i nchange and its necessary.

cEB: example of CFDP in CBHE. well past time on this and continue
discussion on ML.

From Marc Blanchet on chat:
Stephen Farrell said: Not one I care about, but seems like every WG that
specifies standards action for a codepoint range comes back to it later
to change to expert review. yeah. standard action is a very high level.
RFC Required might be more middle ground. It lets an informational rfc,
experimental rfc to get a service number, which is a good level but
still not require full IETF concensus, which is a very high bar.

Delay-Tolerant Management Architecture Updates (10 mins)

Speaker: Sarah Heiner
Document: https://datatracker.ietf.org/doc/draft-ietf-dtn-dtnma/

All updates since 116 are based on review feedback (specifically OPS
Area)

Made changes to focus as an architecture document. Anything not related
moved to ADM document.

Autonomy Engine acts as a Police Execution Engine. Manager can configure
rules in database and enable/disable rules in engine.

"command-based" has been changed to "rule-based".

Feedback to have "event-condition-action" rules.

Conflict from multiple managers

Formal request to move into LC

cEB: will put the LC out on ML. No concerns on the work so far.

DTNMA Data Model (15 min)

Speaker: Brian Sipos
Document: https://datatracker.ietf.org/doc/draft-birrane-dtn-adm/

Cover all then take questions.

Lots of earlier drafts with different names that may have ended up in
implementation. Mkae more familiar and expressive.

There was a lot of rigidity and narrowing. Led to fragility of the
tools.

Missing was handling failures. Reuse the notion of undefined type.

Separating autonomy from core.

Split types to be literal (ARTI processing) and semantic (logical
handling).

Handling time types: Time Point = absolute time using DTN TIme Epoch.
Time Difference is relative.

Clarify processing which was not originally specified. Open to
discussion. Not the right answer but comprehensive.

Due to separation can create a very determistic agent that has nothing
to do with autonomy or care about time

ADM structure in a YANG 1.1 profile

Module processing can reuse stuff in other groups.

Lou Berger: confused by last two slides and what you want to avoid. yang
trasnport mechanism or encoding but not bring along baggage of yang.

What were doing is not incompatability of tooling, existing secondary
behaviors that you wont see on existing tools. Need syntax for ADM
modules and not reinvent wheel of structure text. Reuse syntax and
semantics. When I say netconf ...

Lou: what don't you want from yang?

We don't model and structure data the same as before. All these changes
don't affect ADM or agents.

Lou: we can take offline.

cEB: yes ML. Talked with yangdocs and where does the metadata end and
domain specific language would begin?

Open issues

• what belong in a core ADM?
• in-ADM typing
• protocol between agent and manager

Document edit and feedback. translate ADMs into YANG modules. Trail
implementations?

cEB: please look through this large and important update.

BPSec COSE Security Context Updates (5 min)

Speaker: Brian Sipos
Document: https://datatracker.ietf.org/doc/draft-ietf-dtn-bpsec-cose/

Adopted as WG draft last session. Building on the default context.

Need to have some PKI interop. Use COSE to avoid reinventing the wheel.

Pending WGLC.

COSE ML has general acceptance.

Does not handle policy with BPSec.

cEB: this is in last call, no issues raised in person.

EID Patterns (5 min)

Speaker: Brian Sipos
Document:
https://www.ietf.org/archive/id/draft-sipos-dtn-eid-pattern-00.html

Bundle Protocol YANG Model (15 min)

Speaker: Marc Blanchet
Document:
https://datatracker.ietf.org/doc/draft-blanchet-dtn-bp-yang-model/

management of BP nodes using protocols (netconf, restconf)

BP only using restconf over HTTP/BP.

Can be augmented with contact plans (TVR work)

Looked at existing and attempt to be generic

If node supports v6 and v7 what do?

If node has multiple CLAs?

Looking for WG adopting and comments.

EB: there was work on this on BPv6??. Is there info in prior work that
isn't here?

Don't know.

cEB: we will post adoption call on ML. No say in room.

BP Over Ethernet (10 min)

Speaker Erik Kline
Document: https://datatracker.ietf.org/doc/draft-ek-dtn-ethernet/

BPoE.

May be over direct Ethernet or some virtual Ethernet.

Need dedicated Eth Type and BP version item. No EthType for this
currently in IEEE or anywhere.

Single src/dst currently.

I-D, is this worth it? How fancy?

RT: really like and obvious for use cases. MTU and BP fragment, CLA
should do it. Commonalities for CLs for rules and responsibility for
unreliable underlying transport.

Scott B: long time ago prototype CLA was developed (maybe). helpful to
commonize with CCSDS link layer CLAs.

Lou: ....?

Brian: fragemention is worth looking into. Not a strong model.

Felix Flentge: wondering framentation is really bad, generic
framentation CLA?

selective acks and retraansmit?

FF: some things might be covered by tcp already

RT: one reason i support (lou) and straight to BP and back up again.
less code

Scott B: generic multi-link-layer-framnetation assmebly might be useful

HTTP/BP and SMTP/BP Adoption Discussion (10 min)

Speaker: WG Chairs
Document:
https://datatracker.ietf.org/doc/draft-blanchet-dtn-http-over-bp/
https://datatracker.ietf.org/doc/draft-blanchet-dtn-email-over-bp/

cEB
Application traffic over BP (SMTP, HTTP)
Support for drafts and adoption call

Value in info-doc for application considerations.
Do we have any comments on ...

MB: problem maker. info-doc, prior or during?
EK: AD hat off, IESG hat on someone might ask charter/scope

cRT: absolutely info-doc in charter. if doc built on value, those
http/smtp docs are side effect and can argue where they do.\
EK: they don't need to be adopted

SB: support of having docs. have examples of application being specified
and having something for general principles.

BS: agree with all. http/bp: doing things in bp could make
simplifiations.

Alberto: support applications on network. charter and priorities.

cRT: wg open to people on items based on bandwidth of doing the work.
outer limits of what we can do. charter tool to not do random work out
of scope, not a schedule. see general app considerations is good but
need people to work on it. people willing to work on this?

cEB: general support for info-doc, what do we do with existing personal
I-Ds? work in parallel and talk to iesg on where they go. No dissent

SF: conclusion?

cRT: consenous of room is: support for generic document for application
considerations, no one arguing for adoption of personal drafts but not
out of scope. Marc continue to refresh and align.

SF: reasonable for dtn, some needs to find requirements for these.

Pete resnick: smtp its a gateway from smtp/bp to smtp/ip. gateway stuff
from ancient times and using it here. that would be specific to dtn.

IPN Allocator Identifier Governance Discussion (20 mins)

Speaker: WG Chairs
Document: https://datatracker.ietf.org/doc/draft-ietf-dtn-ipn-update/

Split 64-bits to two 32-bits.

What does it mean to manage such a reg of allocators (upper 32-bits).

Options:

• Expert Review with IANA and IESG appeal
• IESG direct

Who gets smaller encodings? Who gets ranges (and size)?

Can Expert Review meet the requirements for this?

Meeting with IESG mid-aug.
Hold on CBHE Node Numbers.

No comments mean no concern.

Scott B: has concern been expressed already?

Concerns on ML that bias of experts and incosistent rules by them.

SB: how do we consistute this body?

cRT: my understanding is a technical expert not a governance?

SB: breaking new ground?

cRT: rare ground. like ip addr and dns.

SB: how much guidance on how this has been done in past?

cRT: iesg please step forward and give yourself

MB: if ppl care abt what happens under the carpet can be managed. for
deployments instead of protocols and not sure expert review right thing.

Scott Johnson: concern with ER, lots of power of few. need significant
guidance to do correctly. better off with heirarchial allocation so that
default remains with iana and allow the community based stakeholder
governance models (RIR model) to handle where space law is not written
and terrestial law not appropriate.

cEB: large blocks to RIRs does that alleviate the concern?

SJ: yes

cRT: chat with icann how that works. preferred model?

SJ: yes and those existing institutions would be interested.

EK: we can change our mind while writing document. could preallocate.

SF(chat): I left the queue as we're out of time. my comment was going to
be that people seem to me to be getting way ahead of themselves here so
IMO use expert review until that's a problem then if it becomes one, the
WG/community can handle that establishing an ICANN-like supersttructure
for the BP now would be a bit silly

RT: difference between enterprise numbers due to encoding with CBOR.

EK: OID similar.

Open Mic (10 mins)

Speaker: All

Summary and Wrap-Up (5 mins)

Speaker: Chairs

cRT: text suggestions please!!