A A A
Plain text Report abuse
In light of Verizon and AT&T's recent antics in the media regarding POTS and DSL, I thought I might say a word about the phone network as we know it. Strange as it might seem, the current generation of voice switches are a lot more interesting then you might think. Much to the point where you can tell what kind of equipment you're calling (and to a much lesser degree, what long distance network you're using) can be determined just by listening to the sound of ringback from it. Here's a quick recording of a bunch if you'd like to try it for yourself. The difference is subtle, so be sure to wear headphones. The order of the equipment is DMS-10, DMS-100, DCO, GTD-5, 5ESS, and EWSD. There's a couple other things I'll mention later, but more to the point, why is VoIP or wireless such a bad thing compared to the traditional network? Performance and efficiency. IP over ATM, for example, has an overhead of about 9.4% according to . This seems to drive most of the ISP people I've talked to absolutely nuts. But a SIP call by contrast will take up roughly 110 kbps for a call using uLaw. This would normally fit into a 64 kbps circuit - that's 46 kbps (or 71%) overhead per call. Packet loss in and of itself is a whole other topic, but latency is one of the issues that tends to be overlooked. Latency on voice over IP calls can get to be 150 milliseconds (as according to an ITU recommendation anyway. I encourage you to measure what it actually is in practice), something we haven't considered acceptable for most domestic traffic since the days of dial-up. By contrast, an all circuit switched call rarely exceeds 30 milliseconds of latency from one end of the country to the next. As for wireless, the problem resides mostly within the codec. All major wireless standards rely on a method of compression called Code Excitation Linear Prediction to achieve very aggressive compression ratios at the expense of adding a characteristic "underwater" quality. To put it in perspective, Verizon Wireless uses the 4 kbps bitrate of EVRC-B, which employs a vocoder along with CELP techniques, so what you end up hearing is closer to T-Pain's autotune effect then the person you're speaking to. Sprint, AT&T, and T-Mobile (using 8.55 kbps EVRC-B, 6 kbps AMR, and 12kbps AMR, respectively) don't tend to be much better. Even the coming voice over LTE with it's promises of better sound quality rely on just a wideband version of AMR. "But who cares?" you might ask. Maybe you can live with crappy sound quality so long as you understand who you're talking to. That's fine, but consider this; much of the speculation around the interception technologies of our good friends at the NSA seem to indicate that to feasibly intercept and store *all* PSTN traffic, they'd need to be archiving it with similarly aggressive compression. I know this isn't concrete evidence, but if you've ever heard a 911 call when it's released to the media, it's always through something that uses a linear predictive model, and some independent research from another phreak seems to suggest CALEA voice intercepts do the exact same thing. So it's certainly not out of the question. If it's true, the flexibility of uLaw can give us a strong advantage; a layer of obscurity can be added to whatever encrypts your call. It could be analog voice scrambling, a made up implementation of 8PSK with obscured trellis modulation, morse code tapped into a song using a notch filter, whatever. As of right now, the aggressive compression makes this difficult, if not impossible to properly log; there's no packet format you're necessarily limited to. Anyway, getting back to the phreaking aspect of it, the DMS-100 (which is a bit of a marvel within itself. It's current generation of hardware revolves around a redundant pair of Motorola 88k CPUs of all things, and manages to process about 1,500,000 calls in an hour) can give a good demonstration of some of the things the network hides. At first glance, 212-346-9922 is a ringing number. If you're using a phone that speaks uLaw and passes network audio before the call answers, you'll probably know from the sound clips that it's a ringing number on a DMS. While it's ringing, go ahead and make a call to it from another phone. The calls should both go offhook, and you'll be bridged together! Nice, right? These bridges tend to be all over the place, and will hold a good number of callers. Now, if you happen to be calling from a POTS line, there's also a lot of stuff that only you can reach. For example, 958 and your last four digits will get the switch to run a ringback program in a lot of areas. In a lot of former Embarq areas (mostly Centurylink stuff that isn't ex-Baby Bells), 959-xxxx is an internal range that has all sorts of strange goodies; test numbers for IVRs, CNAM readback machines, and so on. A recent 2600 article even covered a bunch of really strange things you can do with carrier access codes - it's like an ongoing episode of Coast to Coast, but with real machines instead of theoretical beings hiding stuff. So - why am I telling you this? I don't expect you, or anyone, to run out and become a phone phreak tomorrow. I mean, the more the merrier, but the last six months alone have been very, very strange times for any politically aware American. I just ask that in between the ever more revealing Snowden revelations, the ever more expanding articles of governments, companies or trade acts pushing for internet restrictions, and the ever more incessant whining of large carriers to regulators to get out of wired telecommunications altogether - voice, data, FTTP and everything, please don't dismiss the phone network as "that garbly thing from yesterday that's probably a series of tubes now". I can honestly say some of the best times I've had in my 23 years have been on the phone network. By writing this, I'm hoping you can eventually experience them too.