On 26/05/2020 14:39, Paul Bone wrote:
ISP A, for example, started up 2-3 years ago and received their /22
from RIPE but now, through growth they need more to service new
customers. They now have to pay a lot of money (in relative terms) to
obtain more IPv4 addresses. To pay for these IP addresses, ISP A needs
to either increase service prices or charge a significant amount for
each IPv4 address.
ISP B is just starting out and has to pay RIPE fees and get on a
waiting list for a /24. All the while being unable to provide IPv4
services.
Whereas large ISPs C, D and E, with their stock pile of IPv4 addresses
don't really care as it does not affect them, and can continue to
charge the same for the service and competition is then affected.
(Assumption made here)
The reality for ISPs A and B is that acquisition of IPv4 addresses needs
to be in their business plan as a cost from day one. There is no way
around this, and arguments about fairness, whilst understandable, are
not going to change that. C, D and E are not, in any way shape or form,
going to start redistributing unused addresses to A and B.
The argument about competition is also somewhat moot - the big providers
can offer a cheaper service by virtue of scale to a degree, and also by
bundling everything together in a convenient £29.99 a month package and
using attractive offers. A small broadband provider is likely not able
to make that fly - so they have to have other value to attract
customers, and not solely compete on price. Yes, that's a big ask if
you're trying to target the mass Internet access market but if you have
other USPs going for you then the relative small increase in cost is
tolerable. If we assume that you don't want CGN (a wholly sane position
to take in terms of initial cost and support woe), lets, for the sake of
making the maths easy, assume that an IPv4 address costs £24 today. If
you cost that at £1 a month per connection, you've amortized that over 2
years which is the default contract length for most consumer access
products these days. You've lowered your margin on the whole deal by
£24, but you still have the IP address in month 25 and it is now paid
for. Now of course, you can't buy them piecemeal like this but worst
case, buying a /24 at a time, you're looking at just over £6k. In the
running costs of even a smaller ISP, £6k isn't going to break the bank
(or if it is, one could argue that the business has other worries).
I've been involved in two ISP startups in the past year - one access and
one predominantly hosting - where IPv4 acquisition costs have just been
factored in as a cost of doing business on the Internet - much like
rackspace, routers, connectivity and everything else. In comparitive
terms, the cost of IPv4 addresses have been one of the larger things to
consider but not something that stopped either being viable. It would
indeed be nice to rewind a decade when we could just say "Hey RIPE,
we've filled up this /19, can we have another one" but that ship has
sailed, docked at several other ports, and is now well on its way
elsewhere.
Your ISP B is doomed to failure if it is sitting waiting for a returned
v4 address allocation from RIPE, unable to offer service: And that's
essentially a business planning/investment failure not an IPv4 scarcity
failure.
Think about it this way. Some of the most valuable spots of prime
London real-estate were, back in the late 1940s, derilict bomb sites.
The people who purchased those for pennies and held onto them have seen
their investment value soar. If I'm just entering the property business
in 2020 should I be complaining that I need to charge my tenants a
higher rent than my neighbours because they bought their freehold when
it was a fraction of the cost and own it outright, whereas mine was
expensive and is mortgaged up to the hilt? It fails your fairness test,
but it is just how it is.
IPv6 is, to some extent, still currently a technical solution for an
economics / business problem - as others have said, the only point at
which v6 will be used in anger is when it is cheaper for large
organisations to do that over sticking with v4. As Per said, the
majority of customers really don't give two hoots what underlying
protocol they are using to reach Facebook - to them it is "the wifi" and
the actual mechanisms are irrelevant. Likewise, any large enterprise
customer is proabably very happy using RFC1918 space internally,
understands enough about routing to make that work between sites, and
don't have a need for IPv6. They are not techies on ukno[tf] - they
will have real work to do, and an employer who won't see the value in
converting the entire corporate network to IPv6 because that's the
latest Intenet addressing protocol. Now some places, if faced with some
major reworking, are indeed adding v6 into their new architecture and
that makes a lot of sense; but these still seem to be the minority.
Most enterprise IT/networking people I know get confused when they open
cmd.exe on their corporate Windows PC and type 'ping bigserver' and see
it using V6 addresses (for sufficiently simple network layouts).
Wearing my consultancy hat, across all of the non-ISP customers I deal
with, only one place stands out as being keen to deploy IPv6 and having
it live on their network.
I don't think there is an easy solution to any of this - ISPs need to
make sure that we offer a good dual-stack solutions to customers so that
when they do want to dabble with v6, or send it live in production, it
works well and they percieve it as a robust and mature protocol. Yes, I
know it is - but if J Random customer enables it to test and it doesn't
work properly, then that customer will turn it off and forget about it
for the next 5 years.
Paul.
