Port numbers and traffic data

[Ian Batten]

>>>
>>
>> These sort of conspiracies were trotted out over alleged plots by =20
>> BT to
>> delay ADSL in order to prop up Nx64 and Megastream revenues.  They're
>> nonsense, simply because BT is too big to run a conspiracy like that:
>> the people who see the revenue increase don't even know about the =20
>> people
>> whose revenue decreases.  In the instant example, selling 10x =20
>> bandwidth
>> at 0.2x the unit cost is 2x the revenue, and if the original profit =20=

>> was
>> 10% it's now 110%, ie 11x better.
>
> Which seems to me to indicate that those dark fibres should be lit =20
> since
> there is insufficient backhaul bandwidth (which is one reason for
> contention) and increasing it would actually make more money for the
> operators. It's just a question of how to ensure that revenue is more
> evenly distributed rather than keeping a dominant operator in a =20
> position
> of power.

But that runs around in circles, because the operator with the most =20
unlit fibre (and the most empty ducting, which is tantamount to the =20
same thing) is BT.  Remember, a lot of the Ofcom price controls are =20
floors, not ceilings.  In the case of IP Stream, for example, Ofcom =20
imposed a price floor in order to energise the LLU market.  I _think_ =20=

the same's true of IP Central, but it's not my space and I'd be happy =20=

to be corrected.  But it's a problem either way. If there's a price =20
control on the floor price for IP Central, BT can only light more of =20
it if they charge the same amount, and that's what happens anyway: =20
they're hardly going to turn away orders.  On the other hand, if there =20=

isn't a price control, so BT can light additional lambdas at a =20
marginal cost of a line card, you're handing the backhaul market over =20=

to incumbents only.

I don't envy Ofcom their task.  The immediate best interests of the =20
customers would be in removing all price floors, imposing price =20
ceilings on a handful of key regulated services and letting the market =20=

sort the mess out.  Which would be fine for a few years, but the =20
punchline would be a consolidation into a handful of well-funded =20
operators, mostly the incumbent, who could cut their prices based on =20
their large volume and drive the smaller operators out.  Then you get =20=

all the sclerotic effects of a small amount of market dominance such =20
as, say, Phorm being able to negotiate with even fewer than three =20
operators to get 85% coverage.
> e.....
>
>>
>> For operators who have their own fibre DSLAMs with 10GigE are coming.
>> In theory you could group 10 DSLAMs with GigE uplink and then use a
>> switch to delivery 10GigE into the fibre, except back in the real =20
>> world
>> there are few, if any exchanges where one single operator would have
>> that density: it would be many tens of thousands of lines.
>>
>
> Yes, one reason I have effectively no choice for my backhaul =20
> provider is
> that the exchange is too small to make it worthwhile for other =20
> operators
> to install LLU kit there. So far at least.

I have a regular argument in the office with my colleague who lives in =20=

the vasty wastes of Gloucestershire.  He complains about the poor =20
quality, and lack of competition, of his broadband, and says that it's =20=

much better in cities.  I point out that the same applies to cinemas, =20=

and no one seriously proposes legal controls on cinema owners to force =20=

them to open a multiplex next door to him.  And, come to that, the =20
same applies to views over the rolling fields: no one's looking to =20
provide any of that here in central Birmingham.   It's bad enough that =20=

people complain about the actions of the incumbent abusing their semi-=20=

monopoly power, but I think you're on pretty shaky ground asking the =20
CLECss to operate out of an uneconomic exchange just to keep the ILEC =20=

honest.

>
> No, but I think the current Ofcom margin squeeze tests are set at too
> high a level. I appreciate that increasing speeds require kit upgrades
> but there must be a way of making the payback periods required match =20=

> the
> investment, at present the payback periods seem to be very short. =20
> Either
> that or there is a lack of planning for ever increasing bandwidth =20
> needs.

Do we want a centrally planned, government run telecoms network --- a =20=

return to party lines, a waiting list six months long unless you're a =20=

doctor and a blanket prohibition on attaching any equipment other than =20=

the operator's to the line, and _still_ lose money in the process?   =20
Bandwidth, like British justice and the Ritz, is available to anyone =20
who can pay for it, and there are generous planning law assumptions =20
about anyone wishing to install it.  If an ISP doesn't have enough =20
between two points, it can buy it from BT, negotiate with someone else =20=

or install their own.  Why should the people who _have_ built their =20
own networks see their advantage (and investment) destroyed in order =20
to favour those that haven't been planning?

>
>
> Perhaps the real problem is that far too many broadband customers are
> expecting to pay next to nothing for their service.

Kerching!  The Phorm argument is that ISPs can't make money at current =20=

prices.  Well, no-one forced them to drop their prices nor, indeed, =20
enter the market in the first place.   And some --- we might mention =20
Demon, or Xen in this --- are (a) not as cheap (b) apparently not =20
talking to Phorm and (c) one gathers profitable.  If ISPs have cut =20
their prices to a point at which they are uneconomic, then that's =20
tough: they're going to go bust, and why anyone should see this as =20
anything other than the correct actions of the market I don't know. On =20=

the other hand, ISPs in fact _haven't_ gone bust, so the claims that =20
they are poor and needy ring somewhat hollow.

But in general terms, purchasers don't have a right to buy something =20
at a price other than that the vendor nominates, and companies are =20
under no obligation to sell at anything other than the price they =20
nominate.  The claim ``we can't get volume X1 at price Y1, and we =20
can't make any money at volume X2 with price Y2'' is the eternal =20
problem of the price-setter: I can easily sell gold in volume at $1/=20
ounce, but not for long, and I can easily make money if I can convince =20=

people to pay $10000 for a cup of coffee, but I mind find the first =20
sale a little tricky.

Peter Houlder charged =A325K/year + the circuit for a 64K internet =20
connection in 1991, and found some takers.  Cliff Stanford charged =A310/=20=

month + the call charges for  a 56K internet connection in 1993 or =20
whenever it was, and again found takers (note the `plus the call =20
charges', which although they didn't initially go to Demon did at =20
least provide back-pressure on excessive use).  Now you might argue =20
that that was niche, and society is in general terms better off with =20
wider availability: fine, but either say you want it regulated and =20
subsidised (and I bet you don't), or accept that using the market to =20
deliver social policy will have problems.

Another argument might be that people who haven't got broadband may =20
only be tempted to have it by low prices, but once they have it they =20
will value it more: my wife has just opined that we'd pay fifty quid, =20=

but we wouldn't pay a hundred, and she wouldn't have said that ten =20
years ago.

Now that's a matter of ability to afford as well as willingness to =20
pay.  But if you read your copy of The Undercover Economist you'll =20
find a dissection of Starbuck's pricing.  Starbuck's would really like =20=

to find out the maximum amount you'll pay and charge you that, but =20
they can't.  What they do is have a range of products at a wide range =20=

of prices, and a lot of add ons (you might be able to buy a coffee for =20=

a quid and bits, but a fancy drink and a cake will cost a fiver).

Contrast the ISP.  I both can and would pay more for my broadband, but =20=

the price has been falling for me along with everyone else.  I never =20
paid for dialup, as we were operating the GBNet Midlands PoP and I had =20=

an account on the Netblazer there, but I think I paid 35 quid/month =20
for 512KB when we/BT first installed the ex-Westell Gen1 kit in my =20
local exchange.  I'm now paying I think 19 for 7.5Mbps.

And ISPs have no obvious way to charge me more.  I could buy a static =20=

IP (why?), I could move to a 20:1 contention business product (works =20
fine for me), but they're not doing a good job of selling vanilla =20
syrup, fairtrade coffee or almond croissants.  Hence, I suspect, =20
Phorm: they see it as a way of getting everyone to order a vanilla =20
latte and pay an extra 35p.  They'd do better to try to find a way to =20=

get more money out of those that can and will pay it for a premium =20
service of some sort, but they don't appear to be willing to think =20
that one through.  I've heard ISPs talk about selling differentiated =20
services at differentiated prices over LLU, ie a richer range of =20
service profiles than those available on IP Stream, but when the =20
rubber hits the road LLU is just used to drive cost out.

We need someone who understands ISP economics to comment, and they =20
probably won't.
>
> I suppose that what we should expect to see if the lowest prices =20
> rising
> while the higher levels stay pretty much static. Reduced contention
> levels would be good too, but my provider doesn't have anything lower
> than 50:1.

Hard to see why: the contention ratio is tunable based on how much =20
backhaul they buy without any need to unbundle.  That's a purely =20
commercial decision.  I don't think ISPs are even constrained to just =20=

the 20:1 and 50:1 BT offer; I believe that if they want to buy a 1:1 =20
match between IP Central and IP Stream bandwidth then, subject to =20
there being enough uplink on the DSLAM (which might be an issue on =20
older kit which only has STM-1 uplinks), they're in clover.  Certainly =20=

in a small exchange I can't see any reason why an ISP couldn't buy =20
10Mbps of IP Central to provision say 20 ~5Mbps IP Stream connections =20=

for a 10:1 contention ratio.  It might not scale to 200 (because that =20=

would be two thirds of the backhaul capacity of a small DSLAM), but =20
there probably aren't 200 customers for that product in anything other =20=

than the largest and most affluent areas.

And look at the pricing for SDSL if you want a slap upside the head on =20=

the real costs of this sort of stuff.  Or the price of 10Mbps of =20
uncontended transit if you're already in the data centre, come to =20
that...

ian