Minutes for ALTO at IETF-91
minutes-91-alto-1

ALTO IETF91 Minutes
Thursday, November 13, 2014
1640-1910
Honolulu, HI (USA)

Chair: Jan Seedorf (present), Vijay K. Gurbani (absent), Enrico Marocco
       (absent)
Note taker: Wendy Roome

Meetecho URL for archived video/audio recording: 
http://recordings.conf.meetecho.com/Playout/watch.jsp?recording=IETF91_ALTO&chapter=chapter_0

-- session starts late because of projector problems (temporary laptop
   from Meetecho team solved the problem, but took about 10-15 to fetch
   and get working)

-- Jan presents agenda and points out that first presentations are grouped
   together and about a general direction the WG needs to decide on with
   respect to enhanced topology representations for applications

-- Jon presents his slides on bigger picture on where ALTO should go with
   respect to topology / YANG / SUPA etc; Richard yang: agree that we need
   to look at the bigger picture; Jan as chair: indeed, the point is that
   the WG should look at the big picture first before jumping into
   particular/specific protocol solutions; Richard Yang: indeed it is an
   open question at what level of detail apps need topology information;
   Rick Taylor: topology information is also improtant for MANET
   environments; Jon: probably not enough energy/people in the room today
   to answer these general questions, we probably need to do this as a BoF
   to get more people involved in this discussion; Spencer: are we talking
   about a non-forming BoF to talk about topology? Jon: idea would be to
   get routing and apps folks together to form a common view on what
   topological data is useful

-- Richard Yang presents on "ALTO / YANG models"; small clarification
   questions, but general discussion postponed after all topology related
   presenations are done

-- Tina Tsou presents on SUPA charter; Jan: are you aiming at working
   group forming BoF soon? Tina: will have breakfast with ADs tomorrow,
   then we will see;

-- Richard Yang presents on "ALTO Topology Extensions"; Jan: before we
   jump into concrete protocol solutions, let's take a step back and have
   a more general discussion on where the WG wants to go with topology
   extensions and what types of information applications actually
   want/need

-- General discussion about the topology-related presentations and where
   the WG should go; Wendy Roome: maybe applications do not need so much
   topology information as we may think; Tina Tsou: from SUPA perspective
   the presented ALTO extensions are on a different level than what SUPA
   needs, this is more for end-user application, is higher than
   SUPA-level; Jan: is this more northbound? Tina: yes, you/ALTO are in
   "arctic" from the SUPA perspective; SUPA is for OMA, needs much more
   details than the ALTO extensions presented by Richard; Rick Taylor:
   let's not confuse the mechanisms and the data that is needed by apps;
   Giorgios Karagiannis: what kind of topology to you want to use in ALTO,
   is it routing topology? Richard Yang: currently we are not talking
   about routing topology;

-- Richard yang: presentation on "ALTO Cost Calendar" can be dropped;
   presentation was dropped as the session ran late

-- Uwe Rauschenbach presented ALTO in wireless networks.  Basic problem
   being solved is whether ALTO can be used to figure out "what is the
   cost to access a server via a wireless cell?" or figure out the level
   of congestion in a wireless cell.  He is seeking to include cell / access
   point information into the ALTO model.  He thinks the primitive 
   connection managers of today (that prefer WiFi over cellular data) will
   evolve into more capable managers in the future and work with the
   applications to work with the changing network topology.  Example:
   opportunistic reaction to handover.

   Wanted to know whether there was interest in moving forward.  Jan
   made a comment that ALTO should not be used for realtime congestion
   control in very dynamic networks.  Jan further noted that we can
   present historical data for congestion control under the current
   charter, but not instantaneous congestion control.

   Rick Taylor resonated with Uwe's proposal since he (Rick) is doing
   this very thing.  Other working groups are also looking at different
   views of this problem and Rick thinks that some aspects of this work 
   may be done in ALTO.

   Someone else (name unknown) on the mic also liked this proposal.
   There is some overlap with SUPA, etc., but like proposal.

   Richard Yang likes the proposal as well.  He notes that the PID
   concept is a partion, not a hierarchy.  Wendy Roome noted that 
   PIDs are upto the server and there can be ways to serve them up
   using differently typed addresses (so far, ALTO has only considered
   IPv4 and IPv6 addresses in PIDs).

   Jan summarized by saying that the WG appears to like this work and
   perhaps discussion can continue in mailing list and proposal be 
   fine tuned as work continues.

-- Wendy Roome presented work in incremental updates.  This work is
   in charter.  The approach Wendy presented designs a general framework
   for continuous and incremental updates for ANY ALTO resource.  

   Jan confirmed that this work solves two problems: incremental updates
   and server-side notifications.  Wendy affirms.

   Wendy goes through her proposal.

   Discussion: Rick Taylor supports the draft.
   Uwe wanted to know why a fixed transport was chosen?  Wendy: Did not
   want the client to choose between multiple transports.  Many questions
   arise if we do that.  Best just used SSE and be done with it.  Uwe
   wanted to know how we will live with broken proxies?  Don't use them.
   Perhaps use a different port than 80.
   Jan wanted to get a sense on whether there were other proposals on
   the table or should we focus on this one and move the work ahead?
   Jan likes this approach (as an individual).
   Jan asked if we split server event notifications and incremental
   update?  Something to think about and take offline.

   [ Wendy skipped her remaining two presentations and gave the floor
     to next presented. ]

-- Lingli Deng presented work on extended endpoint properties in the
   context of privacy-preserving information mapping.  A data model
   has been drafted and is included in the I-D.  Lingli is looking
   for review and some input on next steps.

   Discussion was moved to the mailing list due to time constaints.