except for one: the Interweb scam. The cheapest accounts for both cable and DSL, the ones that most people seem to buy, both have the Interweb problem big-time. However, in the US market at least, DSL offers substantially more opportunity than cable to escape the Interweb scam by choosing among competing wired carriers (Covad vs. the regional monopoly) and competing ISPs. Where cable ISP competition exists, there are fewer ISPs, the competitors form a cartel that ubiquitously implements the same Interweb scam, and the challenger ISPs have not demonstrated that they have a sustainable business nearly so well as these ISPs have done in the DSL market.
Also, I think there are a few germinal laws and regulations that curtail very slightly the Interweb tactics available to DSL carriers but less so for cable carriers. Unfortunately the regulatory environment has always been distracted by a cost-reduction obsession. It has never been more than minimally effective at protecting free speech by limiting the Interweb scam. The DSL regulatory environment could change from month to month, for better or worse, so overall it's probably not worth discussing too much in terms of its direct effects on free speech---only in terms of its indirect effects through breaking ISP competition into small players that don't form cartels.
You must contend with the guy down the street, or possibly the hundreds of other people in your area, for bandwidth. (Andy Smith)I never been able to swallow this as a pro-DSL argument because contending for bandwidth is fundamental to the Internet's design. DSL users also contend for bandwidth---they just contend at a different spot: at the ISP's core network and the telco's ATM cloud.
No matter what your network topology, you are at the mercy of your connectivity providers to provision the network appropriately. Even telco networks---circuit-switched, the epitome of bandwidth reservation, where there is no sharing and no contention---still can have underprovisioning problems. For example, after the California earthquake everyone made paranoia-calls to their relatives, and the voice telephone said ``there are no more lines available.'' And underprovisioned cellular networks like Voicestream's often return ``network busy'' when you try to open a new call. This problem happens all the time in the Philippines, and not just with wireless carriers. There is heavy local carrier competition among wired and wireless carriers, and each carrier must provision separate pools of ``interconnects'' with each of its competitors. You have to worry about through what company your relatives are connected because sometimes calls between a particularly congested pair of carriers won't go through. When Bayantel and PLDT add more interconnects between themselves, this is news that gets published in the newspaper.
If cable gets slow around the after-dinner pornography peak, the proper objection is that your cable company has underprovisioned by making the neighborhood broadcast zones too large, and they need to add more fiber<->copper interconnects throughout their network. NOT that broadcast mediums are fundamentally slower because you have to ``share.''
so, I think the warning that cable's ``download speed'' varies throughout the day and can thus be deceptively difficult to estimate is well-put. I'll certainly agree that far! HOWEVER, I just don't believe that sharing is essentially a scary thing. Everyone who uses the Internet has to share bandwidth with his or her neighbors. Of course it's still important to know if one particular cable or DSL carrier is screwing over their customers by underprovisioning.
There's a related objection that cable is not as well-positioned to add IP QoS as DSL. There's significant technical merit to this position. The merit applies to the upstream direction only, so, for example, if some company were to design a fiber network where a neighborhood of a hundred homes had a gigabit of shared downstream pipe, maybe with link-sharing QoS to insure excess bandwidth is divided evenly among customers instead of evenly among TCP streams (going mostly to P2P users), that'd work. And then said carrier could give 10Mbit/s each of reserved, unshared upstream (TDMA or something). That would work fine.
This objection leads straight into the next one, that bursty upstream bandwidth is fundamentally more expensive to provide on shared-medium networks, thus making the Interweb problem worse. This applies to cellular as much as it does to cable.
but, flitting back to the QoS issue, in practice there is no end-to-end QoS on the Internet. It would be nice to specify how you would like your own packets treated, ex., please give priority to VoIP over ssh, ssh over web, and web over P2P. And better still if you could divide it evenly among family members. To repeat myself, the most advanced kinds of QoS---in particular that latter kind of dividing evenly among family members--are technically impossible in the upstream direction for broadcast mediums, but a ``good enough'' workaround may emerge. I hope so. QoS would be fantastic in the average home or cafe, and is not for sale right now.
What is for sale, is a carrier-mandated QoS. Instead of you deciding which of your own packets get priority, they decide for you which of your packets gets the best treatment. They use this to make a non-neutral Internet. Time Warner probably thinks your music and movie downloading packets ought to have a very tiny priority. Maybe iTunes packets ought to have a much higher priority.
The customer-friendly use of carrier-mandated QoS is to convince you to buy VoIP from the same company where you buy Internet. You can buy VoIP and DSL both from Speakeasy, and it'll be this magical VoIP that keeps working fantastically even when your DSL pipe is clogged full of P2P traffic. Vonage and Teliax aren't able to deliver that, because they would need Speakeasy's help. They can try, but the kludgy QoS they come up with can never be as good as Speakeasy's bundled service.
This level of QoS is provideable on broadcast networks. Both cable and FIOS sell phone service. I believe it is non-IP phone service, so there are QoS-ish ideas involved, but it's not IP QoS. As long as all we want to provide is phone service that needs to be purchased from the same company that's running the broadcast medium, there's no problem providing this, with DSL or Cable.
Personally, I'm able to provide VoIP-carrier-neutral QoS over DSL. It doesn't work great, but it works much better than anything I could kludge to work over cable. I do this by routing all my DSL packets, up and down, through a colocated box in a data center. This is a difference---I can do it with DSL, and I can't with cable. It's a fundamental technical difference: I need to know the exact size of the upstream and downstream pipes to preschedule the packets before handing them to the carrier. A business would have the resources to do this for a branch office, too (and they'd need to since they will have VoIP sessions running between offices, not just sessions running back to their carrier). Business VoIP does not work well over cable Internet connections.
But, in spite of the real technical difference, as far as what's available to a realistic home customer, it's almost the same. You can get good, QoSed VoIP from your DSL or cable/FIOS carrier. Or you can get unreliable maybe good-enough VoIP from anyone.
Cable is one price with huge bandwidth. (Ed Stengel)Not in the upstream direction, it isn't! It's much more expensive, to them, to give you huge upstream bandwidth.
While, from a technical perspective, broadcast networks have this weakness, in practice DSL carriers charge you out the ass for upstream bandwidth, killing any natural advantage they could have claimed here.
ADSL sometimes has "packet" loss, or you randomly lose your info. (Ed Stengel)``your info?'' Obviously we all know that packet loss never translates into information loss on an IP network. It only causes delays, which can mess up multiplayer games and VoIP calls.
I am not yet convinced that DSL or cable has this problem worse than the other. Some broadcast networks do L2 retransmissions to work around normal collisions. Cable might need to do this in the upstream direction, which would add jitter. Jitter isn't the same as packet loss, but it interferes with VoIP systems correctly sizing their playback buffers, and thus has the same effect as dropping packets. I don't have enough experience to say if this problem is real.
If I need a change done to my service, I don't want to have to wade through Comcast's maze of "Do you want to add HBO" cluelessness. (Andy Smith)Good point. I definitely agree that this is worth a lot, no matter what level of guru you are. I don't like any of this bundling.
They're all playing this consolidation, APRU game, though.
Yes, there are limits to how far you can be from the central office to get DSL service. [but...] (Andy Smith)Right now our apartment at 114 Forrest in Brooklyn is not a ``home passed'' for cable, so we can't get cable at all. The TV users all have dishes. DSL can have ``extenders'' if you're too far from the switch. It doesn't matter---they can both extend their coverage as they see fit.
What's more interesting to me is speed upgrade options. In other countries people are getting 10Mbit/s DSL, but I think here in the US our telephone wiring isn't good enough.
Incidentally, Japan is having an ironic problem along similar lines to the Qwest DSL extenders. NTT has wired many office and apartment buildings with fiber, and none of the several competing DSL carriers in Japan can sell to people living in fiber-connected apartments. NTT East is planning to use a wireless protocol called AWA to reach these poor underserved apartments unlucky enough to be on a fiber ring. It doesn't make much sense to me either---why don't they just get some of those repeater boxes?---but that's what I read.