Click the comments link on any
story to see comments or add your own.
Subscribe to this blog
Home :: Internet
12 Sep 2013
I've been having arguments about Network Neutrality with a lawyer.
My position is that you can't adequately regulate ISPs to be neutral, because
there's no agreement what "neutral" means in practice.
He points out that the courts aren't interested in technical details like what packets
are dropped, it's that all traffic has to be treated the same, and ISPs should just
figure out how to do that.
So I contemplated a city with Plumbing Neutrality with the simple
rule that all people must be treated the same
Well, OK, I'm in the commercial real estate business.
I build my building on strictly neutral principles with rest rooms with the same number of fixtures
on each of the ten floors for men and women. All set.
complains that after lunch, she has to wait in line while guys don't.
A court interprets the Plumbing Neutrality law
and decides from first principles that neutral has always meant equal
waits, not equal numbers of fixtures, it's no big deal, just move some
walls. Huh? How am I suppose to pay for that?
I talk to a plumbing engineer who tells me that the rule of thumb, based on the
last 300 years or so of plumbing engineering, is that you need a 3:2 ratio of fixtures
to equalize the lines.
Since there are five fixtures in each rest room, you need to move the wall
to make six in the women's room and four in the men's room. Easy.
But I can't do that, plumbing is heavy so we put the rest rooms on the opposite
sides of the building so they'd be next to the structural walls. Hmmn.
"I've got it," says the engineer. "You have 20 rest rooms, two on each floor,
so you need 12 women's and 8 men's for that 3:2 ratio. So just change
the men's rooms on the 3rd and 7th floors to women's rooms. You'll have to
swap the urinals for something women can use but I can do that."
So I spent more money to replumb and change the signs, all neutral again.
Except there's a guy on the 7th floor with limited mobility, who complains that
the women can just go down the hall, while he has to wait for the elevator,
which takes a while, and that's a problem.
OK, now what? Swap 6th and 7th? What if he gets a promotion and moves upstairs?
My point here is that legal principles are fine, but their
implementation in technology is rarely simple, and the financial risk
of guessing wrong is substantial. Net Neutrality is even worse, since if
you treat all packets the same, your network will collapse, and ISPs will
face endless legal battles about stuff like how much spam filtering is
consistent with being neutral.
So you either need a regulator with the technical
skill to write workable rules, which nobody has, or you need to get the
desired result a different way, such as separating the transport part of
the connection (the DSL or the cable) from the ISP transporting the packets,
as they've done in Europe.
(By the way, the urge to use the phrase "bladder bloat" was nearly irresistible.)
posted at: 12:58 :: permanent link to this entry ::
comments... (Jump to the end to add your own comment)
"all traffic has to be treated the same"
That's not true; that's not what I want. I want:
VoiP packets, to be treated higher than
remote terminal session packets, to be treated higher than
to be treated higher than real-time online gaming packets, to be treated higher than
web-page text packets, to be treated higher than
web-page image packets, to be treated higher than
streaming video packets, to be treated higher than
file download packets, to be treated higher than
ftp packets, to be treated higher than
smtp packets, to be treated higher than
peer-to-peer packets, to be treated higher than
Packet type is what you shape by. That is moral, that is right, that is correct, that is necessary, that is built into TCP.
What is wrong is to shape by address; shaping a competitors traffic over your own.
You treat all packets the same regardless of where they came from.
Not what kind they are.
21 Sep 2013 19:47)
Try watching a Netflix movie on a Comcast connection. They throttle. Yes they do.
21 Sep 2013 19:59)
If you dig deep enough into most of the claims about network neutrality, it seems to me that they boil down to opposing requirements created when connectivity providers are owned by content distributors. Connectivity providers want an end-to-end solution, but also want differentiated traffic so they can make money. Content distributors want to be better than anyone else, but also want to pay as little as possible to get their stuff delivered.
If every connectivity provider didn't have any affiliation to content distributors, then you could talk about having "express delivery" as a business. But as long as Comcast owns NBC, you have a potential problem where content owned by the connectivity provider is charged less than content from another distributor that needs to be regulated. And that is the crux of network neutrality - that like content has like delivery cost and technical delivery characteristics.
There are plenty of political issues that aren't close to being solved, but the addressable technical issue is that neutral differentiation for End-to-End QoS must be set at the origin, not while the packet is in-flight, but if you can't authenticate DiffServ Code Points, you can't keep people from cheating, so you can't realistically permit internetwork DiffServ, so you can't have End-to-End QoS, which means that the Internet isn't neutral today because it doesn't have the facility to be neutral. The closest we can get is network agnosticism, which is effectively 'best effort for everyone', and that is better than gated differentiation when the gatekeepers are effectively or explicitly monopolists.
In practical terms, that means that you marry up MIME type with an IANA registry, modify content servers (principally HTTP servers) to insert a MIME-type code and etag in IP and/or TCP options, and authenticate the header. Then the ISP decides if the header is valid, and if so, the DSCP is admitted to the network, otherwise it is reset to the baseline best-effort QoS level and the origin given demerits.
Ultimately, you'd end up with a mechanism of statistical sampling to validate a particular piece of content is accurately marked, a registry of etags for content, and a reputation web of DSCP markers that would combine to create a more robust way of validating that the DiffServ markings are legitimate than just matching a crypto hash and trusting the good will of the person on the other end. In short, nothing that 10 days of deep technical thinking, 10 months of prototyping, and 10 years of IETF working groups wouldn't fix, but bridge that gets us from where we are to that is supposed to be regulation that imposes penalties for cheaters, and that is also broken.
Also, instead of thinking about plumbing in terms of the people using the water, think about it in terms of the water itself and how it gets distributed to the people. The problem on the Internet is that there isn't enough pipes, not that there isn't enough water, and all that DiffServ really does is create a queueing mechanism that compensates for there not being enough pipes. 'Build more pipes and stop over-subscribing the data network like it is a voice network' is the obvious solution, particularly in fixed line access, but it evidently costs more money than lawyers and lobbyists.
22 Sep 2013 02:15)
Add your comment...
Note: all comments require an email address to send a confirmation
to verify that it was posted by a person and not a spambot.
The comment won't be visible until you click the link in the
Unless you check the box below, which almost nobody does, your email
won't be displayed, and I won't use it for other purposes.
My other sites
Who is this guy?
Airline ticket info
Remembering JD Falk - 10 years later
223 days ago
A keen grasp of the obvious
New Hope for the Dead
465 days ago
Coalition Against Unsolicited Commercial E-mail
Network Abuse Clearinghouse