[e2e] Re: crippled Internet

[Jon Crowcroft]

In message <200104281641.JAA26224 at red.juniper.net>, Dennis Ferguson typed:

the seperate infrastructure argument is a really interesting one - its
true we have several really nearly seperate network infrastructures
running now - now that some of the core IP networks are not actually
sharing the same switching or link infrastrcutre as the core telephony
(or TV ) networks, and we also have wireless telephony and satellite
broadcast tv.....yes we share _some_ edge infrastructure - so this is
where there's soem fun economics - if i want a voice call i already
have it - so to do voice to the edge i na novel way, or to replace a
core, is pretty hard to make a significant business case for - 

also, the core costs for "nailed up capacity" are low enough in places
that really want VPN type tsetups that its not obvious that
multiplexing core switches and links via ip or atm or anything other
than perhapos (perhaps) lambadas is worth the risks (risk to someones
protection from other traffic but more to management errors)...

so its my opinion that (i cant recall who said this but it was i think
at sigcomm 2 years back) IP networks will always offer a QoS tjhat is
"just not quite good enough" for existing services on other seperate
(or near seperate) infrastructure - e.g. classic voice or broadcast
tv, and the integrated service network is a chimera - this doesn't
mean we should nt strive to get better quality on IP networks - but
that the goals should be motivated by existing IP based applications,
service,s and users, and requirements fro mnovel examples of those,
not of traditional existing infrastructure - we have often complaiend
about telco kit being expensive and iniefficient - it was expensive
once coz it was a monopoly set of markets and also coz there wasnt
anything else - we are silly i nthe IP community to expect
the price not to match IP or better (and it nearly does) if IP tries
to compete.... yes TDM is jolly good for voice and near continues tv
broadcast nets...and can be very cheap, simple, robust etc etc

one day, the IP nets will be fast enough e2e that voice will happen to
work pervasively....maybe...but it wont be by direct competition -
more as a side efect of offering other services tha mean ip nets are
deployed to the ede with enough "native" capacity that delay variance
and loss (given new congestio nmanagement techniques and pricing) are
minimized - but we need to motivate this differently i think for
now....

sure there's some voip and ip video deployment - some....as a percentage of
existing billion phone lines and tv viewers i doubt it amoints to a 
blip....it could, but why bother?

now ip control planes, thats worth pursuing

cheers
jon