Quote:
|
Originally Posted by scastle
As has been quoted before, each DOCSIS upload stream is about 2Meg. If you get ten 1 Meg users on one upload stream and they all upload at full speed, then that connection will be fully loaded.
As you may or may not know, TCP/IP requires acknowlegement signals sent back to the server before the server can send the next packet, so if you do get ten 1 Meg users on one DOCSIS upstream, and they are all uploading at maximum speed, then it will slow internet access for everyone else. This to me sounds like a congested connection.
|
believe it or not I actualy understand all this. Yes 10 users (ish) can saturate the upstream. When the 11th user comes on all 11 slow down _equally_ and so and so on as users more user come online. So in your senario each of the 11 users is causing 1/11th of the congestion. No more and no less. My problem is with the idea that if 1 of these users use the connection a lot when it is not overloaded, or his/her use happens to be a P2P app, then they suddenly are creating more than this 1/11th (ie the 5% cause 60-80% of congestion). They cause exactly '1 users worth', no more and no less. Similarly 5% of users can only cause 5% of the congestion. Just because of thier usage in non peak times may be high or of what they use it for does NOT mean they suddenly become able to cause more than 'one users worth' of congestion. See my point ?
Quote:
|
Originally Posted by scastle
As you say NTL got themselves into this mess. Unfortunately,because of the mess, they don't have the money to expand the network. QED.
Having said that, BK is responsible for the mess, so they shouldn't reward him with £3.5K a day.
|
It's a free(ish) market jungle out there. If you make bad commerical decisions then you suffer the consequences. I hold no truck with the argument that users (or the government) should some how 'help out' to correct these problems caused by by commercial decsions. Either NTL are in a position where they can offer competetive BB products or not. If they are not then they should be 'allowed to fail and fail fast'*
* another great Isenberg piece btw at
http://www.netparadox.com/fccletter.html