IRTF OPEN MEETING TUESDAY, JULY 30, 2013 – 9:00 AM

Chair made general announcement regarding IPR disclosure requirements,
links to the meeting slides and announced the agenda for the meeting.

Chair reported that 7 research groups are meeting this week.  This is
7 out of 9.  There is a new research group trying to get chartered on
Network Coding.  It is expected that this research group will be
charted by the end of this meeting.

Of the 9 research groups, 5 have high activity, 3 with little activity
one closed and one that is closing.  Chair is interested in getting
the routing research group back up and active and asked for attendees
to contact him about this issue after the meeting.  Chair also made a
solicitation to the attendees for ideas for new topic for research.

Question:  Mat Ford asked about the ERICAS [Emerging Regions . . . .]
list.  It is not on the Chair’s list of active groups.  This is a
group to discuss issues faced by regions where the Internet is not
ubiquitously deployed.  No further questions.

The Chair then introduced the next part of the meeting.  Two papers
are being presented by two of the winners of the Applied Networking
Research Prize which are awards “for recent results in applied
networking research that are relevant for transitioning into shipping
Internet products and related standardization efforts . . . .”  In
2013, they had 36 submissions and chose 4 papers.

First paper was presented by Te-Yuan (“TY”) Huang a Taiwanese graduate
student from Stanford.  TY presented a paper on video streaming rate
selection.  The paper focuses on the tension between the actual
capacity and the correct rate to fit the capacity.  Many times the
capacity is under-estimated such as picking a rate that is too low.
Rate selection seems to be conservative and this creates problems.
The opposite can happen too in that it can over-estimate the capacity
that can lead to rebuffering.  The tension is created because video
rate is picked based on capacity estimation.  The solution proposed is
to focus on what they know.  For further details, please see the
paper.

Questions:  Wes George from Time Warner Cable.  Question asked and
then turned to Jim Gettys from Bell Labs.  Set of issues caused by
buffer “bloat”.  Video streaming is really “chunking.”  We have a TCP
system that is fighting the video control system.  To what extent have
you looked into these issues?  TY the experiments she conducted
decouple the rate selection from the “buffer bloat” issue.  How can we
make the state of the “playground better”?  This is separate issue.
Yubin Zhao, a student from University of Berlin:   If two competing
flow, how do you decide what is priority?  Nothing the client side can
do but make the best use of the bandwidth they have.  Bill Versteeg:
Don’t just think about the video buffer, remember the TCP buffer.
Question:  Are users annoyed if annoyed when video rate changes?
Depends.  Two things:  what is the rate you are changing to.  Wouldn’t
increasing the segment size solve the problem?  Increasing the segment
size decreases the decision point and you can run into problems.
Chair:  Are other solutions being used?  The sources they checked said
it was a storage overhead issue.

Second paper presented by Laurent Vanbever from Princeton:  Improving
Network Agility with Seamless BGP Reconfigurations.  Difficulties in
changing configurations of network while network continues to run.
Current practices for BGP reconfigurations do not work.    First to
look at BGP configuration from practical and theorectical perspective.


Questions:  Hannes Gredler from Juniper:  Did you look at other types
of technologies are just BGP?  They just looked at BGP.  Olaf (?).
Was it the real production network?  No it was a virtual environment.
Chair:  Is there any hope for BGP?  Is there anything fundamentally we
can do in BGP or is it just fundamentally broken?  Problem is BGP is
so common it is too difficult to fix.  The best answer would be to
throw out BGP and develop something new based on the 20 years of
experience.  There might be techniques to fix and this will just be
another layer of patches over all the other patches.  Juniper suggests
that we may want to consider source routing.

We then had a presentation on MANIAC Challenge 2013 which was held the
27th and 28th of July in Berlin in connection with the IETG meeting.
This year’s focus was a bit different.  Focused on an API fro
recursive packet auctions.  They were studying/testing different
auction strategies.   This year had 5 teams from 3 continents.  More
detail on the website.

General Comments to IRTF:  Elliot Lear from IAB.  IAB is holding a
workshop in December 2013.  Looking at factors that allow work to
succeed and factors that cause the work to fail.   They have reached
out to economist to participate.  They are calling 8 to 10 page
statement of interest.   More information on IAB website.