kwe@buit13.bu.edu (Kent England) (09/14/90)
In article <299@alsys1.aecom.yu.edu> bob@alsys1.aecom.yu.edu (Bob Lummis) writes: >I am planning the installation of an 802.3 backbone in several campus >buildings... > >I am now trying to figure out how to write a bid document, and subsequently >a contract, that will guarantee a good cable plant. > >I could require a series of tests on the cable after it is installed ... > I would not require exhaustive tests of the contractor, since that will drive the price way up, possibly beyond what you can afford, and does not guarantee what you want anyway. I would suggest some spot checks of your own, or by another independent test and acceptance contractor. You need a statistically valid sample. >Or maybe I could specify certain 10BASE-T equipment that should be attached >to the cable plant and successfully used (whatever that means) for a certain >period of time. > Your degradation won't be clearly apparent in all cases. If you suffer a high bit error rate, your performance degrades in nonlinear fashion. The first place you notice it is in poorly designed bootload protocols that perhaps send out huge glops of data in fragmented IP datagrams. If one fragment gets a CRC error, the whole glop is thrown out. After a successful boot, TCP will run "just fine". My point is, protocol degradation in the face of increased BER is protocol specific. It won't be crystal clear when your cable doesn't work well. This may be better than a hard failure, like an open or short. >Or maybe I should specify all of the component types myself and spell out in >fine detail exactly how they should be installed. This is not good enough for >our business people ... > I think you should spec the wire, the punchdown, and the installation technique. Not to mention the closet layout, the outlet layout, and the style of outlet. But that still is not a guarantee that the whole thing will work, but life is like that. I hope I don't get run over by a truck this afternoon, but I can't be sure. I think your only option is to get a LAN installation contractor who will guarantee a LAN as installed. Someone with a contract like that might be willing to guarantee that some particular installation will run 10BaseT. But your wiring subcontractor won't be willing to make that guarantee. He will only be willing to guarantee that the wire you specified is properly installed, not that it meets some technical performance specification. >Bear in mind that what I need to produce is a written document that >will guarantee (legalistically) a correct cable plant the first time You want a pretty open ended guarantee. Where is the liability on the computer manufacturer to make sure the interface cards work correctly? The hub concentrator manufacturer? Are your business people writing contracts to buy 10BaseT hardware with these sorts of open ended guarantees? If so, I want to buy from those guys. But I'll bet they haven't thought about getting a guarantee from the hardware vendors that says it will work on some specific cable plant. Why do they think they can get one from a cable installer? A tailored Cable Plan (based on a plan like AT&T or NTI or EIA) is the place to start. At least you have a prayer of getting a working cable plant, if your installation contractor has some specific guidance on design and installation methodology. Boston University has yet to get burned following this approach of prudent cable plant design with guarantees on the integrity of the wire continuity, and guarantees from hardware vendors that there equipment is "10BaseT". That is what standards are for. --Kent