Notes IRTF #89 NWCRG, London - 6th March 2014

Notes complied from Angeles Vazquez-Castro, Vincent Roca and Victor Firoiu

Meeting started at 15:20.

Co-Chair Brian Adamson was unable to travel. Co-chair V. Firoiu invited V. Roca to help at the meeting from the chairs' desk.

Chair (V. Firoiu) goes through introductory slides on IPR, Note Well and quick overview of the new Network Coding Research Group.

Speaker: Bhumip Khasnabish, ZTE USA, "Impact of Virtualization and SDN on Emerging Network Coding"

- Question by V. Roca: was this work also presented at the IRTF SDNRG group? Any feedback from there?
- Answer by Bhumip: has sent draft to SDNRG chairs and is expecting comments.


Speaker: Cedric Adjih, INRIA, "Experiments with Broadcast with Network Coding".

- Question by V. Roca: what is the packet rate overhead?
- Answer by Cedric: header is less than 20B.
- Question by Bhumip: every transmission is innovative, how is this possible? 
- Answer by Cedric: ? (answer unknown)
- Question by Bhumip: have you considered security issues?
- Answer by Cedric: no


(Remote) Speaker: Raymond Yeung, "BATS: An efficient network coding solution for packet loss networks".

- Question by V. Roca: it seems the object must be known ahead of time, before BATS encoding? Said differently, is it valid only for file transmission?
- Answer by Yeung: yes, but currently developing next version allowing video/streaming transmission.


Speaker: Frank Fitzek, Aalborg University, Codeon, Steinwurf, Phonedeck, "Distributed Storage with RLNC".

- Question by Bhumip: clouds might be using the same physical storage which would create problems. How is this dealt with?
- Answer by Fitzek: yes, they might, need to be looked at.
- Question by Bhumip: what are the APIs being used?
- Answer by Fitzek: the standard ones, provided by each cloud provider.
- Question by Bhumip: slide 18, "number of storage per node" is physical or virtual?
- Answer by Fitzek: it does not matter.
- Lars Eggert: Thanks for putting related IPR info on front page of presentation and on IETF IPR web page.

Speaker: Frank Fitzek, Aalborg University, Codeon, Steinwurf, Phonedeck, "Coding software libraries".

- Question by J. Detchart: dynamic redundancy needs feedback, does KODO allows feedback?
- Answer by Fitzek: no, it's out of scope. The library do not provide protocols, only the encoding/decoding functions.

- Question by J. Detchart: are the codes systematic?
- Answer by Fitzek: no, those are non-systematic codes.

- Suggestion by J. Detchart: in order to perform comparison, a benchmark is needed.
- Clarification by V. Roca: OPENFEC software has two versions, a free one and a commercial, non public one. By default, comparison can only include free versions.


Speaker: Igor Zhovnirovsky, Q Factor Communications, "Dynamic Packet Recovery technology overview".

- Question by Bhumip: is DPR (Dynamic Packet Recovery) physical or virtual?
- Answer by Igor: it is already virtualized.

- Question by Rong Pan, Cisco: to make it end-to-end in a large scale, congestion could not be hidden.
- Answer by Igor: the solution could be to encode little up to the edge and there higher encoding to cope with wireless losses.

- Question by Cedric: is it at transport level, but is it TCP or UDP?
- Answer by Igor: ? (answer unknown)


Speaker: Vincent Roca, INRIA, "Structured RLC codes: an update".

- Question by Angeles: what would be the relation between the signaling and a protocol?
- Answer by Vincent: signaling is essential in practice, and it needs to be considered during protocol design.

- Question by Cedric: does it allow to re-encode at intermediate nodes?
- Answer by Vincent: no, only end-to-end, otherwise signaling aspects become prohibitive in general (although there can be exceptions).

- Question by Victor: what "tuning" are you referring to?
- Answer by Vincent: we select the internal parameters that enable a target performance level and use them later.


Discussion on taxonomy lead by Victor Firoiu

Necessary to bring together people and terminology from information theory, networking and coding theory.

Slides show proposal of roadmap for the NWCRG and candidate activities for 2014. Taxonomy to be finalized by end of 2014 the latest.

Discussion on terminology "Coded Symbol" vs "Coded Payload". "Coded Symbol" seems to be preferred given its use in other IETF WG RFCs such as FECFrame. But "Coded payload" or "Coded packet" have also been extensively used, and are expected to continue being used interchangeably.


Discussion on reference architecture and user cases lead by Vincent Roca 

Slides show a preliminary proposal of reference architecture of NC protocol instantiation based on building blocks, compliant with working methodology in previous (related) WGs.

- Question/concern/comment by Igor, Dan Frost and Bhumip:  not clear what flow is: mixed flows? Same session? Inter-session?. 
- Answer by Vincent: good point, it should be further clarified
- Answer by Victor: it should be further clarified from a coding point of view.

- Question/concern/comment Dan Frost:  maybe a document "Network Coding for dummies" would help other IETF WGs understand the relevance for their own work. 

- Answer by Victor: We were planning for tutorials to be made available at the NWCRG wiki. But a tutorial or introductory document can be done in NWCRG as an Informational RFC. Many others in the room indicated interest and support for such an Informational RFC.

- Question/concern/comment Dan Frost: we need to be careful with decisions that may prevent other groups to make use of the "NC tool". In particular RLC may be hard to process within a hardware component.

- Answer by Victor: this is still a RG, so no protocol specifications or decisions are made in NWC RG. It's not clear whether or not it will evolve into an IETF Working group (as opposed to Research group) or protocol specification drafts will be proposed in other IETF WGs.


Victor asks if the group agrees that the current Taxonomy document on Victor's name can become an RG's doc, and the group agreed without objection. Same question will also be asked on the NWCRG email list.

Finally, Victor thanks Vincent Roca for his help co-chairing this meeting as the NWCRG co-chair Brian Adamson could not travel for this meeting.

Meeting is adjourned at 18:30.