[alto] Minutes from ALTO Session at IETF104

[Jan Seedorf <[e-mail]>]

Dear all,

please find below the minutes from the ALTO session at IETF-104. Many 
thanks to Sabine and Krystof for taking the notes. I have just uploaded 
these minutes to the IETF datatracker. if you have any comments or 
corrections, please let us (the chairs) know.

  - Jan & Vijay

*********************

ALTO WG IETF 104
Prague, Czech Republic
March 26, 2019

Notetakers:
Sabine Randriamasy
Krystof Miziewicz

ACRONYMS
ACAL    ALTO Calendar
DL     Danny Lachos
DC    Dawn Chen
JS    Jan Seedorf
MK    Mirja K¸ehlewind
PV    Path Vector
RY    Richard Yang
SSE    Server Sent Events
TC    Tim Chown
UP    Unified Properties
VG    Vijay Gurbani

ml: mailing list
pls: please

=========================================
TUESDAY 26 MARCH
=========================================
-----------------------------------------
ALTO WG : 16h10 - 18h10
-----------------------------------------

agenda: https://datatracker.ietf.org/doc/agenda-104-alto/
slides: https://datatracker.ietf.org/meeting/104/session/alto
session recording: https://www.youtube.com/watch?v=6LRI0GjpqfA
bluesheets: 
https://datatracker.ietf.org/doc/bluesheets-104-alto-201903261610/
- 17 people in room
- 8 people on Meetecho
Session chaired by Jan Seedorf


===== WG Chairs (presenting: Jan) =====

WG progress is pretty good. Focus on WG docs to finalize.
No new WG items will be added until then.
After that, the WG probably needs to re-charter and find new chairs.
6 WG items are remaining.
Draft dependencies exist: between PV and UP.

Tim chown - JSIC: are there any implementations around ? is there a page 
listing them?
RY: we have an ALTO survey doc with list of deployments and implementations.
SR: for minutes takers asks people to articulate when giving their name 
and affiliation
RY: we have regular meetings with many people, core contributors and 
others interested in re-charter items.
Danny: We have a document on implementations.
Jan: recommends to and will circulate it on the mailing list.

===== (presenting: Sabine) - 
https://tools.ietf.org/html/draft-ietf-alto-cost-calendar-11 =====
- current draft is ver 11 ñ submitted Febí19
- last IESG review in Dec 2018 on version 9
- revision upon review in version 10, this presentation focuses on ver 10
- substantial changes in section 4.1.2: reorganized to clarify text.
- all addressed comments from IESG ware submitted to reviewer. All 
approved. We wait for ART area review feedback (discuss 1).
- Next: we will post revision asap adressing WG Chair review, thank you 
Vijay, and wait for AD feedback

Discussion:
- Mirja: thanks for all those edits. We will have to run another IETF 
last call and put on telechat. I would'nt expect any further comments. 
One quick question: if you use new JSON format would this update 
RFC7285? there was a question on mailing list. What was the decision?
- Jan: afaiu this doc recommends 8259, is up to the WG to update the 
core protocol because 7159 wasn't out there at that time. Sabine, did 
IESG say anything or you introduced on your own?
- SR this was recommended by an IESG reviewer. The WG discuss on whether 
was former WG document tied to other formats than UTF-8. We identified 
none.
- Jan: implementations on 7285 using UTF 8 do not violate the protocol. 
For implementations using this extension (acal) it is clever to use UTF8 
in their core protocol implem. We can stop this discuss because there 
will be another IESG feedback. We feel we are inline.
- Richard: the new JSON format issue is not just applicable to ALTO but 
also to the other docs  such as CDNI. UTF8 format relates to PID name 
encoding.
- Jan: I think we agree. It's not an issue. let's wait for IESG feedback.
- Mirja: then to be on the safe side, please add discussion in the 
document pointing that there are different JSON formats recommended in 
the ALTO WG documents. Would be good to clarify this and were done.
- Jan : Sabine pls note for next revision: the document not only 
requires new format but also explain that other docs may use other 
formats. Then we kick it back to IESG.
- Mirja: I'll put the new doc on the telechat. Maybe nobody will read so 
don't expect further comments.
- Jan: maybe send back to AD commenting on JSON and ask for feedback.
- Sabine: already done. We just wait for B. Campbell's feedback on how 
we adressed his discuss point.
- Jan comment on list by Vijay: everything good, there are just some 
minor comments. Please adress them and we kick it back to IESG.

=====(presenting: Richard) - 
https://tools.ietf.org/wg/alto/draft-ietf-alto-incr-update-sse/ =====
- draft converged after long work
- very important draft. used in ACal. But uses multipart responses, as 
does PV.
- We solved design issue on SSE for multipart responses.
- update 1: incremental encoding
- 2nd major change: compound resources
- Integrated handling of multipart, client ID becomes data-ID
- we seek WG feedback next week

Discussion
- JS: doc is ready now, there were lots of discussions already. I'll 
issue secnd WGLC for 2 weeks.
- Mirja: do we need second review?  - JS: no
- MK: if we want WGLC now => then better issue in 3 weeks. No review 
needed but at least we need some  mails saying this version is ok.
- JS: then we issue WGLC after IETF.
- JS: are there any tech questions? then v17 is ready for WGLC

=====(presenting: Richard) 
https://tools.ietf.org/html/draft-ietf-alto-path-vector-05 =====
- PV: converged after long work
- depends on SSE
- high level topology abstraction
- memory refresh issue (PV cost maps are associated to property maps)
- Capabilities on "co-flows", PV conveys 2 types of data: costmap + 
property map. Propmap allows to compute shared network bottleneck.
- There was a design pb: how combine 2 info ressources in single message?
- design 1: new composite media type. pb = introduces new media type
- design 2 send 2 maps in seperate message: pb = snapshot consistency = 
major problem, see google and amazon.
- design3: introduce multipart/related specified in RFC 2387, including 
2 messages in a single response. Problem: no possibility to know which 
resource is cost map and which is network map because mediatype =  
multipart/related
- other pb: SSE only conveys one media-type. How to deal with 2 media 
types?
note: multipart/related media type = e-mail with attachement
Solution: add the required ìtypeî param in media type in IRD. E.g. type 
specifies type of map.

Next: seek for WG feedback. Chairs will issue WGLC once feedback is 
collected
Jan: Chairs will issue WGLC once feedback is collected. There is a 
caveat: Unified Properties + Path Vector coupling.

=====https://tools.ietf.org/html/draft-ietf-alto-unified-props-new-07 =====
=====(presenting: part 1 Sabine, part 2 Richard)
Uprop as well converged after long work
Current version = 7

Sabine presents problem of ambiguous ressources dependency
Unified Properties for ALTO / part 1
- Revisited definitions of properties
- presented use cases illustrating the problem on consistent property 
map dependencies on resources, inspired from an earlier version of the 
CDNI draft.
2 solutions successively defined. We directly jump to option 2 due to 
time constraints

Richard presents Option 2.
Inspired from relational DBs. Property name binded to dependent 
information resource.
paradigm: table + projection = projection
No time allocated for questions
next: submit final draft end april

Jan: pls read the doc? is WGLC dependent on UP?
maybe wait until both are ready. End of April good schedule


=====(presenting: Richard) 
https://tools.ietf.org/html/draft-ietf-alto-cdni-request-routing-alto-05 
=====
Draft is stable. depends on SSE examples, and on UP.
Issues raised by review not addressed yet.
many security considerations
issue 1: 7285 defines dictionary maps from string -> JSON Value.
CDNI is not a "key : value" store format => cannot call map => shall we 
rename map to table?
issue 2: more generic query:  suggest query by both footprint and 
capability instead of either one or the other.
Next: address other WG comments, wait for WG feedback on options.

Jan: pls send comment on mailing list
WG items done: we are happy WG docs make progress.
we have 25 mins: 3 indiv pres 8 mins each. Happy to see them but no hums 
taken until WG docs are done

=====(presenting: Danny) 
https://tools.ietf.org/html/draft-lachosrothenberg-alto-md-e2e-ns-00 =====
"Multi-domain E2E Network Services"
- objective: feed next charter
- presented to IETF ALTO, SFC WG.
- Reference architecture ETSI MANO NFV architecture
- Collaboration with Ericsson and Telefonica. Look at more partners.

Jan: pls send questions to mailing list. This work is worth being 
considered if ALTO is rechartered. But we need find chairs first...


=====(presenting: Jensen remote) 
https://tools.ietf.org/html/draft-zhang-alto-multipart-02 =====
"Multipart extension for ALTO"
Proposes composite requests and responses.
Compatible with SSE.
Next: find concrete motivatinf example to show workflow and benefit
Jan: please post comments on ml

===== (presenting: Qiao remote) 
https://datatracker.ietf.org/meeting/104/materials/slides-104-alto-alto-for-multi-domain-applications-use-cases-and-design-requirements-01 
=====
"ALTO for Multi-Domain Applications: Use Cases and Design Requirements"
- related to 2 drafts
- Generic framework for multi-domain apps to use ALTO to improve 
performance.
- Define new ALTO design requirements.
- Presntation time was cut due to time constraints.
- Next: get from WG and industrial partners

===== WG Chair wrap up ===============
Next ALTO meeting in Montreal at IETF 105, chaired by Vijay

*********************