[erol]

Originally Posted by ianathuth

The train and the gunman analogy is valid because a few users can affect the service of many.

No it is not. Just as you can only take up as much seat space as the size of your 'bum', so you can only take up as much resource as the the size of your connection. What is more with a connection when all the seat space is taken up the amount of seat space you occupy actualy reduces. In effect when the 'BB train' reaches congestion point _everyones_ bum shrinks equally!

Originally Posted by ianathuth

The two gunmen are grabbing the majority of the seating capacity because of their actions. If ten 1Mb users are online using p2p and maxing out their upstreams (there may be very little us of their downstreams) then they will have saturated that upstream card. The upstream card has a capacity of probably less than 2000 kbps after DOCSIS overheads are taken into account and the 1Mb users at max are using 256kbps. 10 x 256 = 2560 in my book and that is more than the card can accomodate so there is already a slowdown in performance for those ten. Along comes anothr 1Mb user who just wants to download a large legal ISO from a website. He can not achieve the 1048 kbps download that his connection is capable of because his upstream ack packets are slowed down because of the upstream being saturated. Another 100 users come on line at any speed and wish to browse, download or watch streaming video. What will their experience be like? Everyone on that upstream card will have a performance which is slower than dial-up. If the ten p2p users then go offline the other users will more than likely enjoy a normal experience.

I appreciate that because of the total 'bodge' that is Cable BB, upstream usage has a disproportionate effect on downstream (via the need for downstream to use the upstream to say 'hey yeah i got that packet'). What you have described here is a one of the biggest weaknesses of Cable BB btw.

Sounds to me what you want is P2P apps banned. What you hope to get is a connection that shows no signs to you of slow down due to other users usage. OK tomorrow we magicaly 'wave a wand' and on day one your connection starts to improve. Do you really not think that NTL will not just then load more users onto the network until you creep back to the point you started with ?

So what about other applications that use a lot of upstream bandwidth, do you want those banned as well. Like users sending video out from thier machines, or voice, or games or people constantly uploading to their webspace, or accessing data on their machines from remote locations. Do you want these to be 'restricted applications' ?

Do you never upload yourself (apart from 'ak packets caused by your download') ? If and when you upload, do you consider the impact on others both upload and download ?

Bandwidth can be said to be the amount of data that can be passed along a communications channel in a given period of time (look it up on dictionary.com). If the given time is 24 hours then it should be easy to see that a relatively small number of customers on the fastest connections that are using them to the max 24/7 can pass a hell of a lot more data than a very large number of customers that are only on for short periods and may be on much slower connections and just doing a bit of browsing and e-mail checking.

Of course one user can use more bandwidth than another. What matters is how one users usage affects others. Just like you can measure how much road congestion a driver creates by his 'total milage' but only by his milage in 'rush hour', so too is any 'total bandwidth used' figure for a user meaningless in assesing his impact on other users. My problem is not with a statement that '5% of users use 80% of bandwith'. My problem is with statements that 5% of users cause 80% of the congestion.