IETF 116 DTN working group meeting
Tuesday 2023-03-28 00:30 - 02:30 (UTC)
Co-chairs: Ed Birrane (cEB, EB), Rick Taylor (cRT, RT)
Secretary: Adam Wiethuechter (sAW, AW)
Welcome! Notewell. We are recorded it will be on YouTube. Please join
queue on meetecho, even if in person. Mic and video off if not
presenting. Masks on unless speaking please.
Packed agenda! Agenda bash? Nothing, moving on!
https://datatracker.ietf.org/doc/draft-ietf-dtn-bpsec-cose/
https://datatracker.ietf.org/doc/draft-ietf-dtn-bpv7-admin-iana/
Two documents finalizing.
BPSec COSE Context
BPv7 Admin Record Types
Marc Blanchet (MB): when started with RG. There were a lot of code
points not in IANA. Wrote document to register to IANA. We should not do
documents solely for this, must be in spec documents. This should be
last document that does that. All new should be in standard documents.
cEB: please send to mailing list for Brian Sipos. It is also holding up
omething in another WG
cRT: yes last to catch up ommisions in other documents. Totally agree.
https://datatracker.ietf.org/doc/draft-ietf-dtn-ipn-update/
We believe all comments taken.
Fundamentally rewritten to remove rationale text. Replaced with
specification text.
Split into two parts.
However, please review if you made a comment was satisfied.
All IPN URIs look like: ipn:auth-num.node-num.svc-num
Hardest part, dancing around "node". It is logical, not physical. Every
resouce that is colocated on a construct of a node shares a node number.
What was wrong with no authority? Gives discrimation for better
allocation of node numbers. The existing registry and will continue to
be used will use Authority=0 and be the Default.
X taken from existing IANA registry. Forward port of CBHE registry.
Reserved some well known node numbers and maintained existing
allocations.
Node-num=0 is "nowhere". =1 is "local-node" (localhost); non-routable.
2-2^14 is private use (free for all); must assume they will crash unless
in your admin domain.
Snapshot of registry is old in document, will be update.
Feel free to interrupt.
Registered authorities. Org can grab block of node-numbers and use as
they wish and be heterogeneous. New: allowed for range of auth-num.
Subdivision in an allocation is good.
CCSDS could ask range, and hand out one number in that range to
sub-orgs.
AW: is there a mechanism to mark a single for metadata info?
RT: IANA delegates this authority. Can be done but might be messy.
Adminstive issue. Reduce load of iana registry to handle it.
Eric Kline (EK): use now or later? documentation example /31 of
authorities?
RT: final slide but yes
EK: table 3, initial values. No range column in CIDR format.
RT: wary of doing slash notation, we are not an IP address! Felt
weird...
EK: should have range. do you see /notation that needs to be in string
form? are we reinventing class-space.
RT: hard to parse when cidr was introduce. we take this to the list.
0.1.y is localnode only! clarify from chat
CBOR of URI
If default authority continue to use two element, otherwise use three
element. Two element can be handled without a code change.
Three elements incompatible with existing deployments, so interactions
need to be aware.
Feedback!
Reserved auth-num top bit range for other use.
Discussion:
auth-num policy
svc-num policy
formal term for (auth-num, node-num)
EK: call in question for svc-num allocation policy. for resolution
layer! always going to be in endpoint so no need to reserve
RT: ping service number
EK: fully qualified node number
MB: nuetral to proposal. auth-num allocation. IANA does not talk to
orgs. PEM which is a bot. IPv4/6 AS numbers are allocated to RIRs. MAy
want to discuss with IANA from orgs like this. 2. who is expert how you
gonna verify? or asking iana? interesting questions.
cEB: more discussion for list, guidance to desigated experts. one last
slide, please make sure comments are represented.
Scott Johnson (chat):
That is a job for a dedicated entity or group of entities like the RIRs.
https://datatracker.ietf.org/doc/draft-ietf-dtn-dtnma/
DTN-MA draft changes.
v5, high level and solicit final reviews before WGLC.
auto mgmt in challenged environments. cant support request response. use
open loop instead.
self-management with local autonomy agent.
dynamic arch due to dynamic topology
Jorge Amodio (chat): Does DTNMA include sort of a watchdog function?
EB(chat): it allows for it
hierarch model of information
push mechanisms not pull. local to determine what and when to push.
effecient data encoding from data model design
universal data id, so individual datums can be maniped
runtime data defs
auto operaetion
updates to use-cases
Sauli Kivir: comment about notion of remote control and no idea of
availiblity. shared models. approaching change state could lead to
scaling problem. can discuss offline.
SH: can take to mailing list, and explain more.
cEB: further discussion to mailing list but
Stu Card (SC): autonomy, determistic: fully or bounded?
SH: emphasize predicatable and goal is to ensure that when manager and
agent cant communicate we can still know what state applications can be
in that are being managed.
SC: you need to know down to bit or from space?
cEB: good mailing list dicussion.
MB: yang models?
SH: intentinally avoiding anything already providing features. yang is
viable option.
EB: do we hear volunteer to continue work?
MB: looking for yang model of node...
https://datatracker.ietf.org/doc/draft-birrane-dtn-ari/
On behalf of Brian Sipos.
ARI is part of DTN-MA ecosystem. ID and envoke objects in ADM. Out in
couple years with implementations. Some changes based on work.
In work! changes are ARI encodings, not concepts or ADM itself.
ARI syntax simple to use, encode.
Issue address:
simplification
binary encoding is cbor
context-free grammer
Enhanced/simple typing
No compound types for over wire
Real CBOR - now use existing tools and smaller encodings.
Summary
Related Issues
RT: Thanks! havent looked for a long time. ara syntax is about
referencing instaniates members of model and ???
EB: yes.
cRT: parallel work that could be looked at in terms. almost cborpath -
like jsonpath. have looked at any of that?
SH: cant answer for brian. definitely would love to look at existing
approaches.
cRT: go talk to carsten.
https://datatracker.ietf.org/doc/draft-blanchet-dtn-email-over-bp/
BP is not IP! On sale for 2k each.
Rationale
Encapsulation
Consideration
Working on Postfix and uD3TN prototype. Send note.
Steve Farrell (SF): email over previous. could be done. bad idea. when
we did that is was more useful to do sync of message stores instead.
bunch of reasons.
MB: lets argue!!!
SF: pubish mx and smtp over entity??
MB: discuss offline?
SF: layering spam will get a lot. not gonna stop, but wiser to sit back
and think about it. might be message sync.
MB: agree. this way is indepent. sync needs to know format and stuff
thus more implementation details. encap is more standard.
SF: do we want to have this discussion?
cEB: question is what are we asking for? wg adoption or are we asking
way to exchange email over DTN? perhaps we need a statement of problems
and concerns of email over dtn at all. request: take to list.
SF: think that is fair. not an rfc, but think what we want to spend
cycles on.
cEB: queue for people of people in queue. writings about email over bp?
SF: sent to list. can ressurect.
MB: i agree, different ways mailing list!
https://datatracker.ietf.org/doc/draft-blanchet-dtn-http-over-bp/
Now HTTP!
Most things are HTTP based, try to reuse.
Space app dev is a regular Interenet App dev with "best practices"
- far reaching stmt
Encap
whole request in single bundle
uuid before payload for keeping track
Considerations
Comments
implementation in nginx and uD3TN - please send note!
RT: thank you! applies to both smtp and http - more power to be achieved
than just sticking in bundle. destructure logical parts of them and use
multi-block in dtn to store. extension blocks with http headers. chunk
decoding, fix seq? http2/3 fixes the other part. dont just encap. just
like t-shirt.
cEB: longer discussion.
MB: disagree!!!! talk on mailing list
SC: thrilled! unpleasent consequences. Back in the dail was dial on
demond routers, UCCP and did stuff like this - it was how email moved!
still good ideas from those days. to ricks point, lets think about
granularity and layers of abstraction. dont do with fancy webpages but
dont let the endpijnt parse the html, instead proxy does it for you and
sends big bundle. Did it in 1993, can be done but is smart?
MB: can be done using pre-cache and optimation..
SF: less crazy. http client is modified cause it will timeout. we use
HTTPS...
AW: BPSec == TLS??? in HTTPS
MB: ???
SF: app developer, they have security model and it doesnt match here.
cEB: we need to move on (bad joke)
Joshua Deaton (JD): quick comment. like the idea to be taken care of.
not see why we std this specific concept - just do an giant ip wrapper.
set list of app cl. more things to take advantage of! encap is not the
only answer. other things done.
cEB: +1 to open discussion on list
MB: provide text
Zahed: day-0 kind of thinking.
adZ: great that we are talking about it! agree with people are saying -
focus on app we are envisioning and guidance then map to our protocol.
MB: yes
cEB: lets discuss on list!
https://datatracker.ietf.org/doc/draft-sipos-dtn-eid-pattern/
On behalf of Brian Sipos.
define set of eids in a structured way
ipn stuff in reference to updates to said scheme
Use Cases
DTN scheme
IPN scheme
Consideration
cRT: love the work! very important and address EK comments on CIDR.
No Draft
Bakground state research in various areas.
POINTR, cool for space context but fine tune for terresterial
Security is matter of life and death and can not be after thought
variable connectivity and legacy systems
internal implementation of BP, more primitive but concept there...
measurements of what done
lots of wg advice and best practice is needed and useful!
cRT: thanks! cross-over between DTN and DETNET. Ensure that bounds on
characteristics are there. Suggest look at DETNET and other determistic
groupds.
JD: do you have ore info regarding delay resilience of LTP. 4sec delay
range, slightly different stuff...
Feel free to reach out after and share.
cEB: please share to mailing list.