David Kaiser <dkai...@cdk.com> wrote: > Michael showed me the design of his SDSL modem, which was pretty cool, > and we brought up the website [2] over the connection provided by the > actual modem we were discussing.
An addendum to that discussion... David had asked me at the meeting (semi-jokingly I assume) if one can buy one of these SDSL DSU devices I've built. Well, here is the proper answer: * At the present moment, no. In my code of ethics, if I take someone's money for a piece of hardware, I am obligated to provide a certain level of quality assurance, support, warranty, etc. My current rev 0 prototype boards are just that: development prototypes, and although I'm bold enough to have deployed one operationally on my own line (eating my own dogfood), I am absolutely not comfortable with the idea of selling a prototype unit as a commercial product. * I am in the process of revising the board in preparation for making a new batch of PCBs at the next revision level (in the electronics design world we call this a respin), and assuming that these next-rev boards turn out well, the intent is indeed for them to be available as a commercial product. However, when my SDSL CPE devices become commercially available and you want to get one, my advice is don't buy it yourself, instead make your ISP buy it for you, and make them cover the cost! Here is the situation that SDSL service providers are facing: because of the despicable deterioration of the Internet into two polar camps (either you are in a datacenter with a 100 GigE connection, or you are a residential consumer, nothing in between), traditional symmetric business services like T1 and SDSL delivered to non-datacenter locations (offices, shops set up in commercial warehouse space, non-consumer geek residences etc) are very un-fashionable nowadays. Because everyone wants to either move into a datacenter or use some asymmetric consumer broadband technology (ADSL/cable/FiOS), no one is interested in getting SDSL/T1 services, and thus the ISPs are having a very difficult time selling them. Covad (now MegaPath) has an SDSL- serving DSLAM (a D50 like the one in my lab, or at least the LCS part of it) installed in every single CO across USA, and they have the massive ongoing cost of supporting all this infrastructure (the DSLAMs themselves, CO space rental, DS3 or OC3 feeds connecting it all), but I reason that they are probably collecting very little revenue from it, because no one except me wants to use SDSL; everyone else would rather get FiOS/U-Verse or cable, or would rather move into a datacenter. Because of the backwater situation that the SDSL service providers are in, there is absolutely no need for me to hide my activities from them: they are so desperate for someone to buy their unpopular SDSL services that they are perfectly willing to tolerate me using my own CPE device. (I've given demos of my SDSL gear to the network engineers and install techs at Covad and other SDSL operators, so it's not like they are unaware.) But I think we can go a step further: if someone steps forward and volunteers to get SDSL service from MegaPath, we should be able to not only get them to officially tolerate the use of our CPE rather than theirs, but also cover the cost of that CPE as well! All you gotta do is call up MegaPath, say you want to buy their SDSL service, but you have one special requirement: you want to use Harhan brand CPE instead of Netopia, Adtran, Verilink or whatever they'd want to shove to you regularly, and demand that they provide the Harhan brand CPE to you at their expense in order to get your business as a customer. And if the front level sales rep has no idea what you're talking about, 3-way me in onto the call: I would know whom to escalate to. (CPE = Customer Premises Equipment; Harhan = me.) But don't do it now, wait till the production rev of my SDSL CPE is out in another few months. > After standing out in the parking-lot talking for 30 minutes, Michael > gave us the scoop on potential error rates in network cable signaling, > hand-made network cables that could fail, and some other related > discussions, I understand now why Michael prefers 10/100 over Gigabit > networking. Seems like if there is a capacity which unnecessarily > exceeds the minimum requirements, it's a bad thing. I guess if you know > you only need so much, you don't want to have more capacity than that. > A (probably bad) analogy would be: if you only ever needed to drive 20 > mph, and never planned to go faster, you certainly don't need a Ferrari > with a V12 engine and 6-gears. Makes sense to me. So I guess if I > only needed some certain bitrate that was fixed at a few characters per > second, I would want to standardize on 10/100 Megabit gear, since the > extra capacity would just be overkill, and would introduce more > variables or require more maintenance of or more expensive wiring, etc. > It is certainly a personal philosophy, but could be argued to be a good > one for some people. That's a good summary. Of course a lot of people will say that designing your infrastructure only to serve your current needs with no room for growth is terribly short-sighted and a very bad idea, but I don't see how that argument can be applied to Ethernet switches. Because a 100 Mbps switch is really 10/100 and a Gigabit switch is really 10/100/1000, one can easily mix switches of different capabilities. Connect 10Mbps-only devices to a 10 Mbps hub, connect 10/100 devices that prefer 100 to a 10/100 switch with Cat 5 or better cabling, connect Gigabit devices which you actually want to run at Gigabit to a Gigabit switch with Cat 5e or better cabling, etc. With this approach the addition of new gear that likes higher speeds does not impose additional stress upon the older gear and its cabling which remains as it was. MS _______________________________________________ LinuxUsers mailing list LinuxUsers@socallinux.org http://socallinux.org/cgi-bin/mailman/listinfo/linuxusers
