Re: IMPORTANT!! ALL COMCAST Subscribers!! PLZ READ! COMCAST COMCAST

I was referring to subscribed capacity versus actual capacity. I apologize for using an example so outdated (dial-up modems!) that it caused this much confusion.

I wish I could say I don't support it either, but without oversubscription we'd all be paying far more for Internet access than we currently do.

Well, it's _almost_ always a valid assumption, just as the same assumption is almost always valid for POTS phone usage (Mother's Day is one exception) and highway construction, among others.

Fred Atkinson said the following on 9/11/2007 11:50 AM:

If you want to use road analogies, it is more proper to use the analogy of a crowded road with many cars and the bigger ones get picked on first: i.e. trucks are routinely excluded from the fast lane and can be ticketed if they are driving in it (regardless of the speed limit)

Bill wrote in news: snipped-for-privacy@4ax.com:

Over-subscription doesn't make fees cheaper for users so much as it makes more profit for the service provider. I doubt that we'd be paying more, but there'd certianly be less high-speed access available right now. Users have a figure they are willing to pay for a service, and if the price is higher than that ideal price, most of them won't use it. They will wait until the price comes down, and eventually, it will do just that.

"Almost always" isn't "always." It's like pregnancy; it's one or the other. Since it can be demonstrated that a system can be compromised, and since it can also be demonstrated that a system HAS been compromised, providers must face the reality that a system WILL be compromised and plan accordingly.

That doesn't mean business should NEVER oversubscribe, but that the margins should be less and not more. There has to be a plan in place on how to handle the one moment when everyone shows up at the same time. Because it WILL happen.

I am not an end user of AT&T, so I won't worry about it. Comcast is my ISP. Anyway, it will take AT&T forever to develop that kind of packet sniffing. It would have to sort out what is copyrighted, and what is not.

I did? You replied to this:

It didn't confuse me :-)

Absolutely it makes it cheaper for the consumer? Have you priced a simple T1 to your home or business? How about how much it would cost you to get the equivalent dedicated bandwidth (i.e. DS3) that a cable modem can provide you? We are talking as much as a factor of 1000 hear. At one time, I was pulling as much as 26Mbps down and 2.5Mbps up with Comcast until they "fixed" it and I get the standard 8Mbps provisioning. I can tell you that to buy dedicated bandwidth of that sort will cost so much more that just about NOBODY would have high speed Internat at that price.

You don't have to be an "end-user" of AT&T. Your traffic just has to pass through their backbone, and as much as 25% of Internet traffic does.

Here is a little tidbit for you:

All the ratios above are about user modem to isp modem.

Modem ratio is really the Only way that 'over subscription' can be done.

Throughput or the much misused 'bandwidth' is merely a matter of packet congestion.

No they aren't. That value is always 1:1 (considering that with DSL, it is a port on a blade rather than an "isp modem").

Over subscription is simply making X amount of bandwidth available to a given population, P, and the population was sold Y amount of bandwidth, and they are oversold if P*Y > X. If you ever look at the AUP for a service sold under this model, it will always say actual bandwidth levels are not guarrenteed.

No, it isn't. They can be oversubscribe and NEVER see congestion, or a better term is probably saturation. If they over subscribe by too much, or there is abuse, then saturation may become an issue on a regular basis, leaving them the option to either add more available bandwidth or cut the abuse.

Daytripper has never explained how 'bandwidth' can be made available!-)

I don't want to be argumentative, but apparently we just do not see each other's point of view on this thing.

A snippet from the URL you furnished above.

its broadband and dial-up customers, or if it intends to screen all of the packets crossing its backbone network and delete the ones it considers to be copyright infringing. It's also unclear how AT&T is going to be able to tell the difference between fair use -- say someone streaming their home MP3 collection to their computer at work from someone downloading a song from an unauthorized MP3 blog.

So it seems that AT&T is going to be the ones to determine what is copyrighted and what isn't? The above also seems to bear out what I said about AT&T not knowing the difference between what is copyrighted and what is not. IMHO...if AT&T impliments this type of packet sniffing, and start deleting stuff at will, AT&T will be putting themselves in harms way.

Maybe it'll help to do the math. Using round numbers, say you're the ISP and you're offering 6 Megabit (6Mb) plans to your customers and one of your nodes can handle 30Mb of downstream throughput. How many

Next, using oversubscription, (because not everyone will use their service at the same time, and even if they do it only means a general slowdown for everyone), let's say you put 50 or even 500 customers on that same node, resulting in oversubscription ratios of 10:1 or 100:1, respectively. Does that make sense, or did I screw up the math? :)

Thomas and I disagree on many things in this thread, but I think he and I agree on this little sidetrack.

I don't know, but I have used SSL or GnuPG in many cases for many years precisely because of all the *spying* that is going on. Granted, 99% of anything that I do on the internet is unencrypted, but, given the option, I always choose it.

I have read a couple of articles about that AT&T bit and I am pretty sure that there was no question among them [except this article] about whether they plan to use it on the backbone traffic or not; they did. I will just go ahead and assume they are.

Bill wrote in news: snipped-for-privacy@4ax.com:

I agree with this description, anyway.

While "oversubscription" is SOP in resource management, I feel that the provider should take more care in the ratios. 50 people using a node optimized for 30 isn't unrealistic, or even 60, but putting 500 people on that same node is going to result in problems.

It's like an airline: if fifteen people routinely don't show up for their flight, you shouldn't sell fifty extra tickets. There are reasonable calculations, and greedy to the point of stupid ones.