begin oe_protect.scr
Roy Schestowitz <newsgroups@xxxxxxxxxxxxxxx> espoused:
> __/ [ Linonut ] on Saturday 24 June 2006 14:05 \__
>
>> After takin' a swig o' grog, Roy Schestowitz belched out this bit o'
>> wisdom:
>>
>>> I bit off-topic, I know, but it deals with freedom taken away, as well as
>>> disinformatiom.
>>>
>>> Net Neutrality: This is serious
>>>
>>> ,----[ Quote ]
>>> | When I invented the Web, I didn't have to ask anyone's permission. Now,
>>> | hundreds of millions of people are using it freely. I am worried that
>>> | that is going end in the USA.
>>> `----
>>>
>>> http://dig.csail.mit.edu/breadcrumbs/
>>
>> Time to tunnel through the internet.
>
> Grab a shovel, I'll get the dynamite.
>
> In an Internet that is not toll-free, what will be interesting are zombies
> whose traffic is exhausted or whose capturer sells traffic. Like traffic
> laundering...
>
I don't know that it's realistic to expect continuous capacity growth
without some form of management of traffic. I think that there will
emerge multiple networks, which will include the best-effort internet we
have now, but there will come an end to continuous increases in bandwidth
on that, because it's just not sustainable. Somebody has to pay for
the bandwidth, somewhere, the ports, the fibre, the routers, and so on.
With the increasing use of streaming services on the internet, it will
not be possible to support them without a significant redesign of some
of the base protocols. Continuously throwing bandwidth at the problem
doesn't solve it (look at road traffic congestion - users are
fundamentally selfish, and will keep using bandwidth until it's
full/gone), so traffic management is required. Congestion charging is
becoming more common, as are variable speed limits. Just like in road
networks, for EM networks to support all three traffic types, something
has to be done.
Variable speed limits and congestion charging will work for file and
message traffic types, but they do not work for streaming traffic,
because there's no guarantee that an end to end connection can be set up
(a fundamental problem with IP - TCP doesn't resolve it, because it runs
on IP, which has not connectivity guarantee). Consequently, a different
transport mode is required for streams.
This requires a connection-oriented network, much as the PSTN and
leased-line and transport networks of the telcos work now (this
technology is more recent than IP, but is arguable more mature). The
issue is that telco technologies are riddled with all kinds of IPR
restrictions, making them expensive, so next generation networks are
going to run on much more open protocols, but will be able to support
the three networking modes, those being file, message and stream
(everything you do maps to one of these). There is then the interesting
possibility of being able to set up eg., a stream connection for your
remote desktop, say, if you want guaranteed performance, although if you
don't, you could settle for a file type of connection, which would
likely be a connectionless packet (like UDP), mostly fast, but prone to
congestion problems, or a message connection (like TCP), with ack/nack
combinations and proper packet sequencing.
It is my expectation that everything we do now on the internet will
continue to work fine in this environment, but we are looking at an
exciting new possibility, that of being able to really stream video
and audio reliably using stream connections (connection-oriented
circuit-switched) from the same interface (your ethernet port) as you
do for everything else.
I expect that charging models will have to change, but that doesn't mean
that they will become prohibitive - the telcos need the internet as much
as the internet needs the telcos. Telcos are having to learn a new
trick - that of surviving in a commodity environment. This is the same
trick which MS are stuggling with at the moment!
--
| Mark Kent -- mark at ellandroad dot demon dot co dot uk |
Good day to avoid cops. Crawl to work.
|
|
