Flamefests accomplish nothing! I agree with what you say below and at this point not sure what else I can add. Making it happen is what I want to see. I may be wrong but I truly beleive proper UBB it's the only way to true QOS on the internet, especially as applications continue to grow and become even bigger resource hogs. I beleive we previously established that you have a few years on me but this subject reminded me of back in '79 when I worked at IBM as an Engineer. Companies charged for usage of their multi-million dollar mainframes to both their internal and external "customers" in order to keep a handle on resources. The current attitude of internet users is that network resources are cheap. Over-usage is an unsustainable model even with Moore's Law in play. The next few years will be interesting to say the least!
On Wed, Mar 30, 2011 at 12:25 PM, Fred Goldstein <fgoldst...@ionary.com>wrote: > At 3/29/2011 11:15 PM, RickG wrote: > > Fred, I respectfully disagree. > > > I'm glad to see that we have a good discussion going here, and it's not a > flamefest or anything. > > > First off, applications being run on my network ARE my business. Many apps > can have detrimental effect on it and therefore I have a right and > responsibility to say what can run on it. > > > Let's set that aside for a moment... > > > Secondly, priority bits simply cost more to provide and tax the network > more than non-priority. Everyone expects their high priority apps > (video/voice) to be first in line without delays and that's really what all > the fuss is about. > > > I entirely agree. My original note, if you look carefully, said that > priority and other higher-than-BE QoS *should* cost extra. Lack of a good > charging model is one reason why we have a big BE Internet and not much > else. But I'd like to see one. > > BTW, entertainment video is much more tolerant than voice. It can use > retransmission and buffering, or find other strategies to cope. Real-time > video (telepresence) is the bear. > > > Meanwhile, we have been focusing on raw usage but that is only a part of > the equation. Just billing for monthly overages does not consider daily peak > usage times. > > > I don't object to time-of-day pricing either. One rough example: "Usage > midnight to six, not counted. Usage 6AM to 9AM, counted half." > > > In fact, in questioning many customers, they would be happy to pay a > premium for a high-priority, low latency connection for certain apps. Heck, > I can even see premiums for usage based on the time of day but that may be > pushing it. This may sound extreme but everyone laughed at me back in 1997 > when I bought an Allot box for UBB. > BTW:While economic optimization is good, network optimization is better. > Over the years, I've seen fast networks and slow networks, I'd pay more any > day to be on a fast network. > > > Sure. But if your network is capacity-constrained, or the cost of capacity > is particularly high somewhere, then you have to make do with what you > have. Which is why you care what apps are run. > > I am one of those old-timers who thinks that the Computer II decision was > one of the smartest moves the FCC ever made. It literally made the ISP > industry possible, and its revocation in 2005 directly created the "network > neutrality" kerfuffle. CI II banned LECs from dealing in the upper layers; > any "enhanced" services had to be on a fully separated basis. So it was > literally against the law for the telco to ask what the application (in the > "layer 7" sense) was. But it could certainly offer services with optional > QoS features that were optimized for different applications. So it might > offer a CBR service (voice-optimized) and a BE (UBR) service > (Internet-optimized). But if somebody ran voice over a BE service, that was > their own issue, not the telco's. > > On the other hand, the ISP is explicitly not a common carrier, and thus has > the right to be more specific. ISPs are above that boundary, not below. So > sure, as an ISP, you should have the right to sell "applications". There is > no bright line between ISP and ASP, or between ISP and time-sharing service. > (That ancient business has been revived under the moniker "cloud".) So > sure, you can do DPI if you want, provided that you do it in a way that > respects customer privacy. Note that I do not think that DPI should go all > the way in to the payload, as some systems do, or ferret out "value" from > transactions, etc., as some vendors propose. I call that "wiretapping". > > However, I'll quote R. Milhouse Nixon here and suggest that in most cases, > you can do it, "but it would be wrong". The reason is that the value > proposition that customers want from the Internet is one of flexibility and > openness. They probably don't want their "ISP" to be a mail-and-web > service, although that's what Blackberries usually get (in exchange for > unmetered usage of those applications). While I am not one who believes > that "all innovation happens at the edge", I don't want edge innovations to > be blocked in the middle. > > So the best compromise I can see is to price one's menu of services in a > manner that reflects cost, with some averaging to make it more palatable. > And then you can ignore what people do with the bits they buy. They will > have incentives to behave economically. > > I'm also an advocate of widespread encryption, and am doing my part to make > encryption of user traffic the rule, rather than the exception, over the > next few years. (See the Pouzin Society stuff for more details.) That will > utterly break DPI. But it will work with explicit QoS options, so there is > no excuse that the network will have to snoop the application in order to > know what QoS to deliver (the IMS approach). > > On Tue, Mar 29, 2011 at 2:28 PM, Fred Goldstein <fgoldst...@ionary.com> > wrote: > At 3/29/2011 01:20 PM, RickG wrote: > > I still say there needs to be more than just caps. There needs to be a > matrix of billing by priority such as video at .03/meg, file transfer at > .02, email at .01, etc. Heck, perhaps HD can be .05 and SD at .03? (Prices > are just for arguments sake) > > > Well, no, there doesn't. Applications are none of the network's business. > That's one reason why DPI is evil. > > HOWEVER, I am not opposed to appliation-agnostic billing for usage, by > QoS. It is perfectly reasonable for a network to charge for usage that > imposes a cost. And while the teevee fiends are sure, just certain, that > 300 GB/month imposes precisely zero cost on the network, I doubt many WISPs > would agree. Especially rural ones who have to pay for backhaul, or who > have multi-hop networks. > > IP, of course, is one-size-fits-all, with QoS being rare. Hence caps and > overage charges are a way to do cost averaging for the majority (since > people hate billing for usage), while still hitting the heaviest users. > Block pricing (like wireless, having say 10, 50, and 150 GB/month plans, > plus overage) also works. And if you go beyond plain old IP and do have a > QoS-enabled protocol, then lower-loss or delay-limited (or whatever) traffic > should carry a premium. Regardless of what it's used for. Then the > applications could adapt to the pricing. This leads towards economic > optimization. > > On Tue, Mar 29, 2011 at 11:50 AM, Bret Clark <bcl...@spectraaccess.com > > wrote: I know this is Canada, but I can just see some congressman here in > the US one day bitch about not being able to cleaning watch the "Jackass > 3" movie from Netflix and demanding that all service providers get rid of > bandwidth > quotas and throttling by introducing a new bill. > On 03/29/2011 11:26 AM, Matt wrote: > > http://arstechnica.com/tech-policy/news/2011/03/data-caps-claim-a-victim-netflix-streaming-video.ars > > > > > > > -- > Fred Goldstein k1io fgoldstein "at" ionary.com > ionary Consulting http://www.ionary.com/ > +1 617 795 2701 > > > > -------------------------------------------------------------------------------- > > WISPA Wants You! Join today! > http://signup.wispa.org/ > -------------------------------------------------------------------------------- > > WISPA Wireless List: wireless@wispa.org > > Subscribe/Unsubscribe: > http://lists.wispa.org/mailman/listinfo/wireless > > Archives: http://lists.wispa.org/pipermail/wireless/ > > > > > -- > -RickG > > > > -------------------------------------------------------------------------------- > WISPA Wants You! Join today! > http://signup.wispa.org/ > > -------------------------------------------------------------------------------- > > WISPA Wireless List: wireless@wispa.org > > Subscribe/Unsubscribe: > http://lists.wispa.org/mailman/listinfo/wireless > > Archives: http://lists.wispa.org/pipermail/wireless/ > > -- > > Fred Goldstein k1io fgoldstein "at" ionary.com > ionary Consulting http://www.ionary.com/ > +1 617 795 2701 > > > > > -------------------------------------------------------------------------------- > WISPA Wants You! Join today! > http://signup.wispa.org/ > > -------------------------------------------------------------------------------- > > WISPA Wireless List: wireless@wispa.org > > Subscribe/Unsubscribe: > http://lists.wispa.org/mailman/listinfo/wireless > > Archives: http://lists.wispa.org/pipermail/wireless/ > -- -RickG
-------------------------------------------------------------------------------- WISPA Wants You! Join today! http://signup.wispa.org/ -------------------------------------------------------------------------------- WISPA Wireless List: wireless@wispa.org Subscribe/Unsubscribe: http://lists.wispa.org/mailman/listinfo/wireless Archives: http://lists.wispa.org/pipermail/wireless/