Ask Slashdot: AT&T's Data Usage Definition Proprietary? 562
stox writes "As many of you know, AT&T has implemented caps on DSL usage. When this was implemented, I started getting emails letting me know my usage as likely to exceed the cap. After consulting their Internet Usage web page, I felt the numbers just weren't right. With the help of Tomato on my router, I started measuring my usage, and ended up with numbers substantially below what AT&T was reporting on a day-to-day basis. Typically around 20-30% less. By the way, this usage is the sum of inbound and outbound. At this point, I decided to contact AT&T support to determine what exactly they were defining as usage, as their web pages never really define it. Boy, did I get a surprise. After several calls, they finally told me they consider the methodology by which they calculate bandwidth usage to be proprietary. Yes, you read that right; it's a secret. They left me with the option to contact their executive offices via snail mail. Email was not an option. So, I bring my questions to you, all-knowing Slashdotters: are there any laws that require AT&T to divulge how they are calculating data usage? Should I contact my state's commerce commission or the FCC to attempt to get an answer to this?"
Re:Headers (Score:4, Informative)
Is 20-30% A realistic estimation of TCP headers?
If the numbers are correct I would say that a significant portion below the tcp/ip layer is being counted.
How much retransmit/error correction is there in DSL? I personally wouldn't think that's valid to charge, but the argument could be made.
As for for the original poster's question on law, I doubt there is any requirement, though if you challenge them in court, it would have to be revealed, or they have no evidence.
Re:Headers (Score:5, Informative)
DSL ATM overhead (Score:5, Informative)
DSL is based on ATM technology.
And ATM uses 53 Byte cells to transfer data. 48 Byte for the actual data and 5 Byte overhead to indicate things such as the destination.
Now when you want to transfer 50 Bytes of data, you need two atm-cells (vs 1 ethernet packet). This takes 106 Bytes of data on-the-wire.
When one end is measuring the Ethernet side (50 Bytes + ethernet overhead) and the other is measuring the ATM side you will end up with very different numbers.
Re:DSL ATM overhead (Score:5, Informative)
While they may be counting ATM cell usage I doubt it. DSL can use several different encapsulations over ATM. The most common one is a LLC/SNAP header in front of the Ethernet header which adds about 10 bytes per packet. If they're counting cells then the overhead might be higher since the last cell contains an 8-byte trailer containing the packet length and a CRC. Data is broken down into 48 byte cells and if there's less than 8 bytes left in the last cell then another cell is added. It could be even worse if they're using PPPoE which add an additional 8 bytes to each packet, but from my understanding PPPoE is thankfully dying if not dead.
I wrote the data forwarding engine of a BRAS (broadband remote access server) a number of years ago that could terminate tens of thousands of DSL connections. They could be counting cell usage, AAL5 payload usage (ATM frame including LLC/SNAP headers), Ethernet frame usage or IP payload usage.
Contracts cannot be secret? (Score:5, Informative)
I am no laywer and I am assuming the cap is part of your contract with them, I cannot see how they can keep their definition of bandwidth usage a secret. They are now basically claiming that you are restricted in your usage upto the cap but they refuse to tell you what the cap actually *means*. Without clear understanding of how usage is measured, the number of the cap is meaningless.
So you are subject to provions in a contract that you are not allowed to know. It would surprise me very much if they could hold that up in court...
My Usage Matches... (Score:4, Informative)
I had the same problem...once they started charging for exceeding the bandwidth caps I wrote a program to log usage.
I have an old Fedora box with two ethernet cards doing the router work (everything to and from the house goes through this box) and use Etherape to track the usage. A cronjob once a minute makes sure Etherape is always running, and a kill -10 every minute gets it to dump the usage data in XML which I process into a CSV for analysis and charting.
Surprisingly, their monthly usage figures have matched my full month calculations within 1%.
What irritates me is that their monthly totals are not available on their WWW site for a full week after the end of the month, and their current month totals are also delayed a couple of days sometimes wildly inaccurate since they are missing days. Example is the November totals for my account seem to be currently missing 2-5 November, and they haven't posted 12,13 November yet. Hence they show lower usage than what I really used. If this were the end of the month, I might think I can squeeze that extra download in before the end of the month, but I am sure they would figure it out and charge for it.
I hit this issue once when I breached the 150Gb cap with 6 hours remaining. They claim to sell you another 50Gb for $10, but of course that doesn't roll into the next month. That is where I would complain....if they are going to charge by the Gb, they need to accurately report usage during the month.
AT&T just sent me a letter that they are switching me to U-verse with a 250Gb cap. They claim it will be the same price as DSL for the next year, but after that who knows....only other game in town is Comcast which cost even more.
Re:1k increments (Score:4, Informative)
That is incorrect. Ablock specifically works by blacklisting URL patterns from being requested. I don't know exactly how noscript works, but it's surely going to stop a script from requesting other scripts or ads.
Re:1k increments (Score:2, Informative)
Completely wrong.
Both adblock and noscript prevent the browser from fetching unwanted content.
Re:Headers (Score:5, Informative)
Yep. And they do.
That's why I go elsewhere.
That being said, my mom used AT&T. They never got her bill right (always overcharged), and we had to call and they would correct it, but having to do that every month or two sucked. I told that to an AT&T solicitor at my door once, as to why I wasn't interested. Rather than defending his client and get a sale, he responded, with a dismayed and somewhat depressed, "wow... I heard about the same thing from someone just down the street," and moved to the next door. My mom doesn't know anyone on my street other than me, so it seems to me, overcharing is not an uncommon issue with them, and they are just trying to find more clever ways to hide it.
The author of TFS should verify that he's collecting headers, if so, he may have a case.
Re:Headers (Score:5, Informative)
For a 20% to 30% discrepancy and claims of proprietary measuring system screams one thing and one thing only, random sampling and rounding up. Basically they are averaging out usage and rounding up in their favour. Basically going for lie, cheat and steal until challenged by a class action law suit forcing openness and accuracy.
Re:Headers (Score:5, Informative)
Unless someone is sending an awful lot of really small packets, the 40+ bytes of TCP/IP headers per packet are not going to add 20-30% to the data that is being sent. For example, the "Simple IMIX" as defined on WIkipedia has 58% of packets being 40 bytes long (they are common because they represent data acknowledgments with no data going in the other direction), probably significantly underweights the number of 1500-byte MTUs, and still only has ~12% TCP/IP overhead. It would be grossly inappropriate for AT&T to include any packet overhead beyond TCP/IP because any lower level overhead is an artifact of AT&T's network design that is outside the control of, and opaque to, the end user.
Re:Headers (Score:5, Informative)
Re:Headers (Score:5, Informative)
Sure bandwidth costs may increase as usage increases, but so what.. they are charging for it.
They would love you to think that, but actually bandwidth costs have decreased significantly as usage has increased. The problem is not external bandwidth costs, it is oversubscription. They don't want to invest in expanding their internal network as much as they need to, instead just lumping more and more users on the same local pipes and hoping their access patterns are all little bursts. Of course streaming and P2P kill that.
Re:Headers (Score:5, Informative)
Used to be more true than it is now. Most fuel station tanks in the USA are 2-3 ft underground, below the frost line, so the stored fuel temperature stays at a relatively constant 50 or 60-something degrees even on the hottest summer days. Sure, if it's a bloody hot day at a station that isn't used much, the fuel that's actually in the pump may warm up a little, but they retain very little gas.
Unless, of course, you lived in Centralia, PA. Then....then you have a very good point.
Not in Canada (Score:4, Informative)
In Canada fuel sales are volume corrected (to 15C) so regardless of the temperature of the fuel you pay for the same amount.
Re:Headers (Score:4, Informative)
Cost is not the issue.
Aircraft measure fuel by weight because that is what is important to flight.
How much weight am I lifting?
How does the weight the fuel and the placement of the tanks they are in affect my CG?
How much more fuel do I have to burn off before my weight is within limits for landing?
This is why Military, Commercial and General aviation use weight.
Re:Headers (Score:4, Informative)
How are they doing that? That's the entire problem. They're saying 'how' is proprietary and they won't tell you.