[erol]

Originally Posted by andygrif

I think you're misunderstanding me.

If I am it is not on purpuse and I appologise for it. I am getting a bit confused when you refer to 'bandwidth costs' but I am not sure we undersand the same thing from this.

Originally Posted by andygrif

What I mean is that if you double the size of the potential capacity, you would have to assume that, even sporadically, more actual data will be consumed en masse.

This will cost ntl more money, this means the concept will not work.

It will cost NTL more money if they have to increase how much external connectivty they buy to support users, which is a reasonable assumption. My point is that this increased costs in _minisucle_ when compared with the overall cost of providing the service. I would estimate that external connectivity accounts for around 1% or less of the total cost of provision for the service. Thus if NTL doubled eveyones connection speed overnight (and assuming that _thier_ internal network could handle this) and that everyone then double their data volume usages such that NTL had to buy twice as much external bandwidth then the overall impact on costs would be neglible, as the cost element that has doubled (though prob less than double to double their external connectivity) is tiny to start with. Then add into the equation the fact that the cost of external bandwidth is falling year on year fatser than the cost of processing power is falling.

The fact is, if external connectivty was a significant cost then you just would not see retail services at only a few pounds more than the BT wholesale IP Stream product, which only pays to get data from customer to ISP and nothing else (ie not external connectivity). The fact that many many companies can operated on BT wholesale prices + a few pounds more to cover everything else (external connectivity, staff and office costs, profit etc etc) is about the simplest clearest 'proof' that external connecivity is a tiny fraction of the cost of provision.

Originally Posted by andygrif

Also, I don't see how providing an email server (or two) can cost more than the data transfer from ntl's provider.

I would suspect that NTL pay at least as much or more for such services as they do for their external connectivity. You have machines / hardware for email servers, newsgroup servers and proxy servers. You then have software licesense for all these. You then have staff costs to configure and maintain these servers and staff costs to support the problems they cause for users. The hardware costs are falling about the same rate as the cost of external connectivity. SW is not falling so fast and staff costs are just going up (bar outsourcing to other countires). So even if NTL do not pay more for running their 'servers' than they pay for external connectivity today - they soon will do.