Rex Ballard <rex.ballard@xxxxxxxxx> espoused:
> On Mar 12, 6:56 pm, Roy Schestowitz <newsgro...@xxxxxxxxxxxxxxx>
> wrote:
>> Study: Net neutrality law would spur infrastructure improvements
>>
>> ,----[ Quote ]
>> | One of the main arguments opponents of network neutrality make is
>> | that laws requiring ISPs to treat all traffic equally would
>> | discourage infrastructure investment on the part of those who own
>> | the pipes. A new study conducted by researchers at the University
>> | of Florida undercuts those claims.
>
> Keep in mind that it was net neutrality which greated the demand for
> high speed bandwidth in the first place. If the telephone companies
> had had their way, we would still be trying to squeeze 28,000 bits/
> second out of a dial-up modem, or even more likely, would still be
> trying to squeeze 1200 baud out of an accoustical coupler.
Verizon are currently building a fibre-to-the-home network, which
involves a massive amount of digging up roads, installing new equipment
in street cabinets and LEs. The big issue with the local loop is that
it is extraordinarily expensive to install, and then to maintain, so
getting affordable high-band loop to domestic premises was awaiting the
technical steps forward to make it possible, and also the kind of
services which would result in demand to make it economically viable.
ADSL was the enabling technology, whilst the internet was the demand
source.
>
> The divestature of AT&T resulted in diversification of technology.
> There was the transition from analog 24 pair cables to copper cables
> that could carry 1.4 megabits/second over each pair, to Fiber that
> could carry as much as 5 gigabits/second. Video phones had first been
> suggested as a possibility back in the 1950s, and AT&T even had
> technology to do the Analog-to-digital and digital to analog
> conversion of video signals, along with various forms of compression
> even as early as 1980. The problem was that Cable and Telephone used
> different bandwidths and different circuitry. Today, digital cable is
> not substantially different from that 20 year old technology - first
> suggested as a possibility by Brian Kernigan and Dennis Ritchie.
Okay, but 140Mbit/s (in Europe) and 45Mbit/s (US) plesionchronous
transport were commonplace in carrier networks before the web was even
invented. The hierarchy in EU goes 2-8-34-140, in the US, it went
1.5-6-45. There was also an interworking hierarchy, but that's not all
that relevant here. At 140Mbit/s it was quite feasible to digitally
code broadcast TV to transport it around the country, and this certainly
happened in the UK. The bandwidth requirement is huge, though - a
140Mbit/s path can carry 1920 telephone calls, or one encoded PAL
signal.
>
> The irony is that carriers, both cable and network/telephony have
> provided grades of service. When a business wants to do video
> conferencing in real-time, they pay a much higher price, or a premium
> tarriff for the faster bandwidth required to get smooth video. Cable
> companies are the same way. The broadcasters who want smooth signals
> containing music and action movies delivered smoothly, the cable
> companies deliver - for a price.
If you look at the cost/bit as the bandwidths went up, you'll find they
were less expensive rather than more so. However, the issue with good
video streaming is that you cannot do it effectively on connectionless
packet networks - you need connection-oriented packet switching, merged
at layer 2 with the connectionless stuff, not at layer 3, so that there
is no collision between streams and files and messages. Look here for
more info: http://en.wikipedia.org/wiki/Carrier_ethernet_transport
>
> Flat rate IP service has always been available, but there are also
> different tiers of service, including commitment rates, quality rates,
> and availability rates, as well as other options such as fixed IP
> addresses. It's entirely possible to purchase the same cable, with
> the same wires, for as little as $29/month and as much as $299/month,
> depending on the quality of service you want. Most carriers even
> offer higher tiers of service, for those who want to host an entire
> subnet or run a hosting service.
The contention rate is a fairly key thing. Business contention rates
are often down to as little as 10:1 or 5:1; domestic ones can be as
high as 25 or 30:1, depending on how much you pay. This is independent
of the local loop technology (be it dial-up, isdn, adsl, leased-line or
whatever).
>
> One of the big fights over IPv6 has been the desire of end users and
> customers at the termination end wanting to have control over their
> own subnets, with the ability to route and/or tunnel to the correct
> server. The carriers, on the other hand, want to have control over
> every IP address and charge for each independently. Furthermore they
> want to have the ability to partition the IPv6 subnets by geography,
> meaning more wasted IP addresses, but the ability to charge by
> distance as well as quality of service.
Distance-based charging has been standard in telco land for much longer
than flat-rate charging - there are pros and cons, though, since complex
price-lists make it difficult for customers to make comparisons.
>
> One of the biggest problems lately, is these new services that allow
> you to watch your own TV - over the internet. The problem is that
> this is personal use of about 4 megabits/second, to carry what amounts
> to redundant content.
How would that be redundant?
>
> As both a cable subscriber and a frequent traveller, I can understand
> the frustration of spending $99/month for cable, including premium
> channels like Cinemax and Showtime, then spending 20 days/month at a
> hotel where there is only one channel of HBO (if that).
>
> What would be nice is if I could take the access card out of my cable
> box, put it in my wallet, and when I get to my hotel, put it in their
> cable box, and get the access ot the service for which I'm paying
> triple digit service fees.
But you'd need one company providing all kinds of access for that to
work, and also, there is the very practical point that /someone/ has to
pay for the hotel's local loop. The argument you're making here is
essentially the pro-IMS argument, which means that you need to be able
to authenticate yourself from anywhere by secure means, and get whatever
content you want delivered to that location. Furthermore, the content
servers have to have some means of understanding just what network
you're on the end of, and what its capabilities are (is it a mobile
phone, say? A wireless PDA/tablet?). Additionally, unless one company
is running the whole thing, the content needs to be distributed across
multiple companies' networks, with you being able to access a remote IMS
system in order to authenticate yourself.
I just cannot see the above really being practical. A more viable
solution would be to have at least quasi-independent service offering
which you pull down wherever you are, and merely authenticate yourself
with the server, but pay for the transport to whatever your location is.
If you happen to be in a hotel, though, they could charge you a /huge/
amount of money for doing that.
>
> Instead, the media industry wants us to pay premium rates for home
> service we can't use, which includes the same 8 movies on 8 different
> channels staggered by a few hours, even though we have VCRs or TiVo or
> DVRs.
If you have a TiVo, why don't you put a few things on a DVD or a
flash-card and play them on a PC, GP2X or Nokia 770 or something like
that?
>
> Then they want us to pay $4.95 per movie for "pay per view" to get a
> movie on demand. Then they want us to pay $4.95 per movie for
> "download on demand" movies.
Hotels want to charge you for whatever they can, though...
>
>> | Researchers at the school used game theory to come up with an
>> | analytical model for a world without net neutrality. The study
>> | argued that the killing net neutrality would result in fewer
>> | infrastructure investments and service upgrades rather than
>> | more.
>
> The counter to this is competition. If cable and DSL have to compete
> with MIMO and 802.11n relay centers to the backbones, or 10 mb/s
> cellular, the nature of competition is that all carriers will try to
> compete for various stages of the bandwidth.
Only if there is money in it. So far, there has been a complete dearth
of wideband services which anyone wishes to pay for; and even the IPTV
services being launched are getting around the
connectionless-packet-switching problem by downloading the /whole film/
before you play it. Ie., they get around the buffering problem by
having a buffer which is as big as the whole file, and then move it as a
file instead of a stream.
>
> By 2001, one of the factors that caused the "Internet Bubble" to
> burst, was that the carriers had built up too much excess bandwidth.
> They could sell it to each other, but the unused bandwidth just became
> an expense (which caused some confusion for Worldcom who was treating
> bandwidth like commodities futures).
Worldcom had lots of other problems, though. It should be noted also
that leased bandwidth traditionally had very long lease times (>25years)
so losing old capacity is not as simple as it looks. The dot-com bubble
burst wasn't really much to do with the telcos, it was much more about
all the VCs who'd invested in startup telco equipment providers, on the
basis that the internet would somehow make money and they'd get some
cash back. All the major NEPs got stung in this, resulting in much
consolidation which is still going on now, this includes Alcatel buying
Lucent, Ericsson buying Marconi, Nortel???, Nokia buying Siemens
telecoms, and so on.
>
> Many businesses now use gigabit WAN connections. Some of the media
> companies now use terrabit networks, often needing high speed switches
> into huge "trunks" of distribution.
>
> The problem is that a popular video can actually flood the internet.
> When Bill Gates video testimony was put on the web, streaming video
> systems were pumping huge volumes out, because the video was cached,
> but the network routers and switches were struggling to keep up with
> the bandwidth.
The internet is just rubbish for broadcasting... it's much better to
broadcast using radio networks, be they terrestrial or satellite, since
you don't need to start up a new stream to every single end-point. The
way the internet has traditionally got around this problem has been by
using local mirrors of popular material. The old sunsite network being
one example, many academic institutions offering the same thing. In the
more modern era, the likes of google do their own mirroring/clustering,
which helps, but even then, if /everyone/ wants to see the same thing at
the same time, setting up an individual connection to a server from a
user for every single user is /never/ going to scale.
>
> The price of network traffic has gone from 40 cents/megabyte to 4
> cents per gigabyte. The "flat rate" service has created the demand
> for these services, but the "premium rate" service has financed it's
> development and expansion.
What traffic would that be?
>
>> | Even worse, ISPs that require content providers to pay
>> | for preferential access on their networks run the risk of losing
>> | customers whose favorite sites aren't on the preferred list.
>
> Keep in mind that this is not a block-out. It's only a traffic
> shaper. Traffic shapers and priority routing have been available on
> CISCO routers since the early 1990s. In the early days, this was
> needed so that FTP traffic didn't make telnet users have to wait 10
> seconds for an FTP packet to complete before the echo came back.
The real problem with traffic shaping is that it can only be used to
prevent some things happening, but when the network goes into the
"meltdown" state, it's pretty much a useless thing to have. You really
do need deterministic networks to handle super-demand, so that you have
a concept of "busy-tone" which is much better than thinking you've got a
link up, but actually getting zero throughput.
>
>> http://arstechnica.com/news.ars/post/20070311-study-net-neutrality-la...http://tinyurl.com/36ej68
>>
>> This parallels Free software in the sense that limitation and lockin give the
>> vendor no incentive to improve.
>
> This is based on a very "binary" point of view. It's based on the
> monopoly mind-set. The problem is that the only monopoly in the
> entire internet, is the Microsoft desktop. From the first switch out
> of your house, to the final destination, and back, every switch
> competes for routing priority. The main difference is that the PPV
> video coming into your cable box has a higher priority than your web-
> cam chat with grandpa. The reality is that you probably won't notice,
> because your web-cam only uses about 256 kbits/second, even at 30
> frames/second, and a 200 millisecond delay merely means a slight
> "jump" or "jitter" in the bandwidth.
>
Unfortunately, connectionless packet networks are just not that
predictable in performance. When you consider just a couple of streams,
it's not so bad, but consider what happens in the core of the networks,
where there are meshes of routers forwarding packets which are part
of hundreds of thousands, if not millions, of different streams.
The probability of *your* packets suffering a problem is much greater in
such an environment, because the network is in a chaotic state at times
of high demand very similar to the state big motorways get in just before
they go to synchronous driving mode (where noone can overtake since the
whole road is full). However, since both motorways and IP networks are
not connection oriented, a slight chance to traffic at that critical
time brings the whole thing down; in the case of a motorway, you get
those queues which can build up for miles with no apparent cause at the
end (studies have shown that something as simple as a driver touching
his brakes can be enough to cause a 30 minute, multiple-mile queue in
peak time).
Fundamentally, you cannot guarantee that your packet can get onto a
buffer, so when things are bad you are likely to lose packets.
Running streams over such networks will never work well.
>> Just look at the sorry state of Internet
>> Explorer 6, which turned the Web experience into a nightmare while offering
>> no further development for 5+ years.
>
> Yes, and because Microsoft dropped the ball, competition generatod
> Opera, Mozilla, FireFox, Galeon, and Konqueror. And the popularity of
> Firefoox led to adoption of OpenOffice, which has increased the
> popularity of Linux and Unix (OS/X) on the desktop.
>
> Microsoft still maintains an illegal monopoly on the OEM channel,
> demanding exclusionary contracts which are in direct violation of the
> Clayton act, but at this point, due to the available of low-cost high
> speed internet access, including WiFi at Starbucks, Panera, and IHOP,
> CableModems, and DSL, it's as easy to download quality applications
> for Linux (including Linux itself) as it is to install it from a CD-
> ROM or DVD. I downloaded a SuSE Linux DVD ISO image starting just
> before I went to bed. By the morning, I had the image, and burned the
> DVD in about 15 minutes. 10 minutes later, I was rebooting my PC on
> that same DVD, and installing SUSE Linux on my Z61p.
>
The internet has been a truly disruptive technology in so many ways,
this is just one of them...
> I still have the OEM license to XP, and I still have the ability to
> install XP as a secondary VMWare client, but I don't have to limit
> myself to Windows alone.
>
>> This should be demand-driven, not capacity-driven, where capacity is
>> artificially limited, with imposed barriers.
>
> The irony is that nobody noticed the limits before, but they have
> existed even since the early 1990s. The problem is very similar to
> what happened with Napster. Someone looking to make a quick buck
> selling a service which didn't seem so bad when it was only a few
> hundred users, has expanded that service to 10 million users, who are
> calling their ISPs and complaining because their streaming video to
> their work computer isn't as smooth as they want it to be.
Well, you can't do realiable streaming on IP... at least unless it's a
point to point link, anyway.
>
> They are upset because their $29/month service isn't getting the same
> priority as the corporate video conference being provided to a
> customer who pays $20,000/month for 5 gb/sec connections to 20
> locations.
Corporate video conferencing would typically be carried on leased-lines
or ISDN lines which are connection-oriented, so that the quality is
guaranteed. The problem with these technologies, though, is that they
are very expensive, because the vendors charge *enormous* amounts of
money for them. One reason why PBB-TE (see link above) offers such
interesting possibilities is that it could readily be set up using
ordinary linux machines, although you'd still need a separate ethernet
which could drop into it's own VC (or VT for Sonet) or even wavelength.
If you were doing a corporate campus version, then wavelengths would
probably be easiest.
> The low cost service promises a minimum commitment rate,
> but does not assure smooth delivery of that rate. The low cost
> service has a maximum rate, but that maximum rate of say, 4 megabits/
> second, might be 1/10th the minimum commitment rate. If there is as
> Starbucks down the road, and one of their customers, who is paying $10/
> hour, decides to download a DVD, his traffic takes priority over
> yours, simply because the Starbucks has a connection that might have
> 50 people connected to the WiFi hub at $10/hour per person.
Starbucks would probably have a business contention rate, though,
whereas your domestic service probably won't.
>
> If you want the faster performance for an hour or two, go to
> starbucks.
>
:-))
--
| Mark Kent -- mark at ellandroad dot demon dot co dot uk |
| Cola faq: http://www.faqs.org/faqs/linux/advocacy/faq-and-primer/ |
| Cola trolls: http://colatrolls.blogspot.com/ |
|
|
