Presence and IntComm WG

[Jeremy George <>]

Hi Jeff,

  Good thoughts.  Thanks.  I have a little different notion, or maybe
just some follow on notions.

>While use of low bandwidth codecs and codecs that are less susceptible
>to packet loss (e.g. iLBC) are helpful, this will not be adequate to
>solve the problem of all users sharing a 802.11 connection of the nature
>established in the Indianapolis Marriott.

  Absolutely.

>I also think voice is an integral part of "integrated communications".
>If 802.x wireless technology is to be used in such real-time application
>demonstrations,  QoS metrics need to be established or such
>demonstrations will certainly suffer.

  The problem with this is that general deployments of QoS (beyond p-p
links, etc.) just aren't going to happen any time in the near future.
Do we really want to say that VoWiFi can't exist without QoS?

  Indy was kind of a special case.  The WiFi net was experimental,
exactly as our demo was.  It was a learning experience :)  What we
witnessed was tantamount to congestion collapse, and there is more
we can do than just adopt iLBC (although that's a good start.)  As I
noted, see draft-iab-congestion-01.txt .  It's possible that real time
traffic over udp contributed substantially to the problems but I doubt
very much that VoIP was a major factor.  While the lessons from the iab
are directed toward VoIP, I think they apply equally well to all real
time communications.

  It seems clearly true to me that voice is integral to IC; I'm not sure
where the question comes from?  What I'm not so sure about is whether
voice in the context of IC needs to be five 9s reliable.  I'd be willing
to accept something less than that if I had seamless integration with
other media.  The analogy I'd make is to cell phones.  Great feature set;
not five 9s.  Ditto IC.

  So, I think there are two issues.  Congestion collapse & what can be
done to prevent it.  And the need for Five 9s for voice.  According to
the iab we can do quite a lot to help prevent the former.  I guess the
latter is a spiritual issue until we know more about how people will
behave with P&IC.  I think this might be an ideal opportunity for some
paths-in-the-snow engineering.

  I do also agree though that conference demos might represent worst
case scenarios.  If we can make P&IC work at them, it's highly likely
to work in a more normal context.

- Jeremy

-- 

----------------------------------------------------------------wg-pic-+
For list utilities, archives, subscribe, unsubscribe, etc. please visit the
ListProc web interface at 

    http://archives.internet2.edu/

----------------------------------------------------------------wg-pic--