[e2e] Is a non-TCP solution dead?

[Cannara]

Well, David.  I wonder what those links between peering nodes are; what those
metro links to service providers are, etc.  I don't simply mean leased Tx
lines, which may have confused you.  I mean telco-standard equipment, as you'd
but from any of the major vendors, like Juniper, Cisco, Turin, Mahi, Nortel,
Lucent, Avaya...  Those devices are generally fiber linked these days, and
though the traffic is moving to Enet framing, the dominant structure that
supports the near ends is essentially Sonet, whether it's GbE over Virtual
Concatenation, or simply STSxxx.  The point being made is that the telco-style
network, datalink and physical layers must meet telco standards and generally
use the same telco chips and underlying management to deliver these services
to metro customers.  The links to customers are again, telco standard, up to
and apart from RF.  And what, pray tell, do we think links RF towers
together?  RF for local nets is, of course just that, local.

By the way, Christian's suggested reading in PDF is here:

http://citeseer.nj.nec.com/cache/papers/cs/3129/ftp:zSzzSzsunsite.unc.eduzSzpubzSzpackageszSzinfosystemszSzmbonezSznvzSzgreed-j.pdf/shenker94making.pdf

Alex

"David P. Reed" wrote:
> 
> Ah, most excellent Cannara - I'd be interested in some quantitative
> evidence for your assertion that the Internet runs almost exclusively on
> well-groomed Telco lines.  Almost all of the Internet traffic I experience
> runs over LANs on part of its journey, just as most of the Internet exists
> close to the edge.   I personally suspect that most end-to-end traffic will
> be carried via RF and optical photons in a few years, copper will be used
> exclusively for power distribution, and fiber will be viewed as an
> optimization relevant to the minority of communications that need never be
> mobile.