Home page logo
/

interesting-people logo Interesting People mailing list archives

Re: A series of interchanges as a result -- How a few inches become 500 miles.
From: David Farber <dave () farber net>
Date: Sun, 21 Sep 2008 18:30:16 -0400



Begin forwarded message:

From: "Bob Frankston" <Bob19-0501 () bobf frankston com>
Date: September 21, 2008 6:20:23 PM EDT
To: "'George Ou'" <george_ou () lanarchitect net>, <fin () finseth com>, <nnsquad () nnsquad org >, "'Dave Farber'" <dave () farber net>, <oia () lists bway net>
Cc: <Jason_Livingood () cable comcast com>, <tkoppel () comcast net>
Subject: RE: [ NNSquad ]  How a few inches become 500 miles.

George, I’m not sure why you are trying so hard to defend the status quo when it is so obviously flawed and is within the context of a marketplace architecture that is so dysfunctional. If you could improve some of the current protocols and business models what would you do – is today’s system the best that we can have?

Perhaps I’m all-too-aware of the problems because I was around as the stuff evolved and still think of it as a series of fortunate hacks that have accumulated far too much sludge and have been forced to work within the context of a 19th century theory of telecommunications.

Note that I explicitly address the issue of logical vs geographic distance. The protocols still work but the real world aspects affect performance and we also are forced to run past a billing point.

As to cell phones – why are you an apologist for the current design even if it is fatally flawed? If you’re not near a tower you can’t communicate. Yet the phones are likely to both have 802.11 (though we should be able to do far better than 802.11). It would be very hard to make the current system more complex than it is – if you look at the protocols they are very complex because of the billing requirements and the channel allocation requirements. Don’t confuse your naïve design assumptions with what can be done.

You’re right that the routing between two points tends to be static – so why do we act as if the network has to dynamically track all the end points. Why don’t we learn from the post office and use relatively static addresses for routing and relatively static names for naming. That would make the protocols far simpler than today’s protocols which are all optimized for the most dynamic case and make it difficult to take advantage of any of the stability. And they fail in their task because there are no stable names and today’s network fails if you disconnect parts from the root!

From: George Ou [mailto:george_ou () lanarchitect net]
Sent: Sunday, September 21, 2008 17:54
To: 'Bob Frankston'; fin () finseth com; nnsquad () nnsquad org; 'Dave Farber'; oia () lists bway net
Cc: Jason_Livingood () cable comcast com; tkoppel () comcast net
Subject: RE: [ NNSquad ] How a few inches become 500 miles.

Bob, you’re conflating geographic distance with IP distance and that’s a mistake. There’s a good reason that you’re not meshed in with your neighbor who’s on a different ISP which in the IP world is as if he was in a different city. No ISP is going to have mesh peering with another ISP with a million peering points and millions of potential routing loops.

As for cell phones not being able to peer with each other directly, that’s a design decision to keep the system simple. For the vast majority of the time, you can’t assume that two cell phones will be within an earshot of one another and for the few times you’ll be able to do cell-to-cell communications, it just isn’t worth adding that complexity to the whole system. It’s also imperative that that cell phones be able to communicate with the tower so that they can be scheduled to use the radio without fear of collision. The last thing you want a cell phone network to be is an unmanaged wireless network like plain old Wi-Fi with thousands of ad hoc devices transmitting on the same channel anytime they please. This is why your unmanaged Wi- Fi network using 20 MHz of spectrum is only capable of 4 simultaneous phone calls while a managed cell phone network can do hundreds of calls in half the spectrum.

“The whole point of the Internet is that these aren’t special cases that have to be managed individually. I shouldn’t be required to use path-dependent protocols. And the whole point of telecom is that I must.”

Routing between any two devices on the Internet, contrary to popular belief, is generally very static and predictable. At most, if both end-points are dual-homed with BGP, you might have 4 different paths to choose from but even then the other 3 paths are only used if certain paths fail.


George

From: Bob Frankston [mailto:Bob19-0501 () bobf frankston com]
Sent: Sunday, September 21, 2008 2:20 PM
To: 'George Ou'; fin () finseth com; nnsquad () nnsquad org; 'Dave Farber'; oia () lists bway net
Cc: Jason_Livingood () cable comcast com; tkoppel () comcast net
Subject: RE: [ NNSquad ] How a few inches become 500 miles.

George, you’re missing the point. I know I can wire around this problem. And of course I do. But be careful of excusing the status quo just because one can work around some of the problems some of the time. It’s an example, it’s not a problem I’m trying to solve in itself.

The problem is that if I go to my neighbor I can’t do this. I can’t watch the town hall meeting if it’s on one silo and I’m on another lest I use up my quota and/or get degraded performance.

There is indeed no reason to forward my packets between me and my neighbor via New York. So why must I?

And there’s no reason that if my son is using my Comcast connection he has to do anything special to use the printer on my Verizon-connected home network. Of course I can make that work too with enough effort. The problem with provider-assigned addresses much deeper and is one we’ve been wrestling with since the Domesday book in England a thousand years ago – how do you have identifiers that are independent of location and path.

The whole point of the Internet is that these aren’t special cases that have to be managed individually. I shouldn’t be required to use path-dependent protocols.

And the whole point of telecom is that I must.


From: George Ou [mailto:george_ou () lanarchitect net]
Sent: Sunday, September 21, 2008 17:06
To: 'Bob Frankston'; fin () finseth com; nnsquad () nnsquad org; 'Dave Farber'; oia () lists bway net
Cc: Jason_Livingood () cable comcast com; tkoppel () comcast net
Subject: RE: [ NNSquad ] How a few inches become 500 miles.

“Today there is a real problem. I have both Comcast and Verizon. To get from a Comcast port to my small FTP server six inches away but connected to FiOS my connection travels through NY or NJ and loses 90% of the performance. A very long six inches.”

Bob, I would suggest that you have an engineering problem of your own making here. Rather than blame Comcast and Verizon for this intra- home networking design flaw, redesign your home network to use Fast Ethernet or Gigabit Ethernet between devices subnets and use a router that can intelligently route within the home network between your home networks and two ISPs using the shortest path. There’s no reason for you to be forwarding your packets to either Comcast or Verizon if you’re trying to travel between subnets within the home.


George

From: nnsquad-bounces+george_ou=lanarchitect.net () nnsquad org [mailto:nnsquad-bounces+george_ou=lanarchitect.net () nnsquad org ] On Behalf Of Bob Frankston
Sent: Sunday, September 21, 2008 12:01 PM
To: fin () finseth com; nnsquad () nnsquad org; Dave Farber; oia () lists bway net
Cc: Jason_Livingood () cable comcast com; tkoppel () comcast net
Subject: [ NNSquad ] How a few inches become 500 miles.

As Craig notes “Unfortunately, traffic from your local city hall meeting may well travel halfway across the country and back before it gets to you.” Thus we should replace the requirement of "community access TV" with the requirement of local peering. But why not just get rid of the whole warped system?

Today there is a real problem. I have both Comcast and Verizon. To get from a Comcast port to my small FTP server six inches away but connected to FiOS my connection travels through NY or NJ and loses 90% of the performance. A very long six inches.

If a city selects Comcast Transport as its connectivity provider you can be sure that you'll get strong local connectivity and peering with nearby communities. CT would be a contractor maintaining the community’s common facilities and would bid against others for that contract. I doesn’t make sense to cede rights of way to a myriad of silos. That’s just like laying a separate track for each trolley company along each street. At least tracks would be visible and create outrage at the waste of space and money.

Distance doesn’t matter to the protocols but it does matter for performance and, far worse, it matters if the only reason for the complexity and overhead is a dysfunctional market model embodied in physical hardware.

Forcing bits to travel long distances is due to having Comcast Transport and Comcast Content fused into the a silo and doing the same for Verizon FiOS (and RCN and Verizon copper and ATT Cellular and T- Mobile and Verizon Wireless and Sprint and the MVNOs …). You have to transit within a silo to some distant land where they agree to exchange bits held hostage at a high price in money and in performance thus destroying value for no reason other than sustaining a market architecture that would not exist were it not protected from the threat of abundance (again – http://www.frankston.com/?name=AssuringScarcity) .

We see this with the again and again. Put two cellular phones one inch apart – the voice/SMS bits cannot get between them unless the signals can reach distant towers. Actually for some bits it’s just the opposite! You have to put two phones near each other to exchange information using IR or Bluetooth – strange when you have radios that can provide connectivity without regard to distance. It’s as if someone picked up an obfuscated C book and assumed it was guide to best practices.

I’ve got Vonage and ATT VoIP as line 1 and line 2 on one phone. Again – the bits travel the long way for no reason other than to create a billable event (or justify a very high flat rate). It’s all about billable events and nothing but billable events.

The current Internet protocols aid and abet this by being path- dependent rather than end-to-end. I get an IP identifier from a provider’s silo rather than generating my own (except for local-only addresses from my NAT which again give me different kinds of bits). Forcing the identifier to also act as a path-address reminds me of the 1960’s when we’d put disk addresses in database records “for efficiency” or use octal disk addresses as user names (as on the PDP-10).

As our financial system suffers why are we trying so hard to assure profligate waste and costly hurdles by insisting on a horribly inefficient system that prevents us from creating our own solutions and takes what is already paid for and assures we can never get the benefits of local control and ownership. Why do we want frustrate the creation of new value?

Telecom – the real financial debacle. It’s like a mortgage that is 100% interest and at a variable rate that can never ever be paid off.

Yet I’m told there is no political will for change? Is madness our due?

-----Original Message-----
From: Craig A. Finseth [mailto:fin () finseth com]
Sent: Sunday, September 21, 2008 09:54
To: Bob19-0501 () bobf frankston com
Cc: Jason_Livingood () cable comcast com; tkoppel () comcast net; nnsquad () nnsquad org
Subject: Re: [ NNSquad ] Re: Comcast's FCC Filing Today

            ...
One question is whether the cap will apply to local connectivity as well as distance. Is there a limit to how long I can watch a broadcasts of local city hall meetings? I presume the answer is "yes" and that it is a policy
   question.
            ...

Unfortunately, traffic from your local city hall meeting may well
travel halfway across the country and back before it gets to you.

Not a problem if you live in a major interconnect point (bay area,
Chicago, New York).  A big problem if you live in <somewhere>,
Nebraska.

This brings it around to the public policy side as well: depending
upon where the encoders are, the cable company may also bring the
video traffic along the same path.

Craig





-------------------------------------------
Archives: https://www.listbox.com/member/archive/247/=now
RSS Feed: https://www.listbox.com/member/archive/rss/247/
Powered by Listbox: http://www.listbox.com

  By Date           By Thread  

Current thread:
  • Re: A series of interchanges as a result -- How a few inches become 500 miles. David Farber (Sep 21)
[ Nmap | Sec Tools | Mailing Lists | Site News | About/Contact | Advertising | Privacy ]