Minutes: IRTF RRG (7 Nov 2005)

Taken by: Dow M. Street

Presentation 1: Avri

Avri gave the status of the RG and introduced Dan Massey, the new co-chair of the group who should be able to help in reaching out to the research community

A the last meeting the work of the future routing domain scalability sub-group was presented. Since then

- extensive update sent to the list within the last month (in archive)

- interesting discussions on list

This RG group is looking at research project that both

• long range and fairly speculative and

• medium to short term that are closer to having IETF routing area relevance

• BGP stability project is still a pending idea if there is sufficient interest in this.

The general RRG list has been quiet, but picked up slightly in the last few months. This is an open list and can serve as a good forum for testing the waters in a wide area of routing topics.

Currently trying to get a RRG meeting at some of the academic conferences

• There will be an informal meeting at Infocom in Barcelona.

• if there is enough work to warrant , RG can meet at IETF more than once a year. Aaron F. is working to have Friday afternoons open for research group discussions

Reviewed status on Requirements for inter-domain routing:

draft-irtf-routing-reqs-04.txt

• main issues have been taken care of

• Yakov's comments on the goals and reality of EGP to be added as editorial

• document is out (as well as history document)

• another pass is planned and then to have a "research group last call"

• won't change the content of the original groups, but will clean up / add notes

Presentation 2:

Elwyn Davies - Analysis of IDR requirements and History

draft-irtf-routing-history-02.txt

Taking an evolutionary approach to the document

• trying to keep the doc going because it captures lots of useful info

• will integrate some of the newarch project

Bob - how mix the history with the requirements?

Avri – it is a history of inter domain routing requirements

Aaron - the GIG BOF is developing emerging requirements - these are out of scope, right?

Elwyn - it is history, and documents problems, but will not include future / emerging requirements

Presentation 3:

Dan Massey spoke on routing convergence

A lot of work is being done in academia and the ACM / IEEE research community thinks there is a convergence problem

• work has, however, had no current impact on existing protocols

Issues:

number of flaps affects the convergence time

- more timers involved

- after enough times, the routers close to the event do the damping and things get better

There are a number of proposals from research community..

- MRAI timer

- assertion checking

- ghost flushing

- root cause notification

Is routing convergence an issue to take up in the RRG?

- sub-area to try and track / move this into protocols

- more a question for IDR?

After some discussion a tentative 'yes" was reached

Dan - these are still a little far from IDR

Avri - there are common folks in this meeting and IDR

? - criteria to use to judge relevance, but what about a different internet?

not a problem now doesn't mean it's not a problem later (circumstances involved)

Ross Callon. - chicken and egg, vendors will not get excited unless customers want it or there is a real and deployable approach - make something work better in a way that leads to more money

- can we get the researchers to discuss on the list?

Bob - should the RRG work on X? well that depends on the other things it could be working on (what are y, Z, and W?)

Presentation 4: Bob Braden

Newarch / routing in the future internet

- Dave Clark was a principal contributor (and Xiaowei, now at Irvine)

- timeframe of 10-20 years out

www.isi.edu/newarch

ideas:

- economics

- trustworthiness (user doesn't distinguish between security and robustness)

- greater heterogeneity

- ubiqitous mobility

- expanded application architecture

- meta-architecture

FARA global network does not require global address space

NIRA routing arch for user empowerment

others (in slides)

NIRA - source and dest address describe an implicit up-down route

other details..

for RRG, is user empowerment a good idea, in general?

- reason we have telephony based - end user route control

- customer only pays upstream provider, not the next upstream

- similar to big argument at NANOG (multi-6) - this level of the TE is up to the providers

- Shim 6 on the verge of re-inventing this work

- does the IT dept at a campus (with multiple providers) want end-users to decide?

- Key idea: look at source address to determine which path to use (in multi-homed hosts)

Bob Briscoe

compensation should follow the choice

also allow concurrent paths

- 2 competitions (price and form / structure of tarrifs / etc)

- this allows different business models to arise

- in NIRA, if you keep to pairwise contracts, each node can act as a buffer between business models

- if allow more meshed, it requires the standardization (get more notes on this)

- perspective from an IETF WG chair

- how can IRTF be useful?

Any other IETF folks present who can comment?

Bill Fenner - routing AD

- to some extent, everything we're doing is engineering because we're not doing the things that aren't

- existing groups are scoped to well understood problems

- on the back burner, a different architecture for inter-domain routing

- keep hacking BGP, how long will it keep going?

- providers have reasons to keep current model, but if (ex. addressing) comes as a part of a whole system then change is more possible

Sam Willer - have we learned enough from v6, or do we need more work on transition?

Could we move to something like newarch, or what specifically (related to transition) do we need to work on?

Bill - v6 hasn't taught us what we need in that it is the same addressing and routing arch

- taught us something, but not much about addressing or routing transition

Avri - how one evolves to a new routing architecture? getting folks to accept anything new is hard - idea of evolution itself is interesting

Unnamed: for radically new paradigms, CIDR is a good example of experience here. Hopefully new arch doesn't need to go outside the IPv6 address space

Dave - i've noticed lot of penetration of DHT style protocols into routing - is RRG a right place to dig into this?

Dow Street mentioned the problem of GIG mobile routing (to be discussed at the GIG BOF later that night

- basically the macro-mobility problem between the MANET work and the Internet model

- another idea, multi-path routing randomly src-dest, TCP does all the TE (more traffic on the better paths)

RRG list is an open list for speculation, argument, discussion

when IETF folks run into a research sort of problem, send it to the RRG