ietf
[Top] [All Lists]

Re: IAB Statement on Dotless Domains

2013-07-12 09:11:03
    > From: Phillip Hallam-Baker <hallam(_at_)gmail(_dot_)com>

    > for many years it was IETF ideology that NATs were a terrible thing
    > that had to be killed. A position I suspect was largely driven by some
    > aggressive lobbying by rent-seeking ISPs looking to collect fees on a
    > per device basis rather than per connection.

That is so confused.

First, many (the majority?) of people in the IETF who didn't like NATs had
sound technical reasons for so doing (it breaks end-end, makes third party
referrals in peer-peer applications harder, etc). Those of us who diagree
with them don't (in general) disagree about those costs, just think the
benefits of NAT outweigh them. (See below.)

Second, while the ability to have a per-device fee might have seemed like a
nice fantasy to some ISPs, the reality is that their costs are driven by i)
the total amount of bandwidth used at the site, and ii) the costs of
providing the connection (hardware, configuration, etc). Anyone who tried to
monetize per-device would have had competition from people who only charged
based on their actual costs. And given that NATs are so easy for consumers to
set up, I think most ISPs realize they save them a bundle in customer support
costs (given that each customer call costs them some amazing amount of
money); the inevitable support costs from per-device would diminish the
amount of money allegedly to be made.

    > From: Keith Moore <moore(_at_)network-heretics(_dot_)com>

    > NATs still are a terrible thing that need to be killed. They break
    > applications and prevent many useful applications from being used on
    > the Internet. That much is more widely understood now than it was 10-15
    > years ago.

You're still ignoring what _empirical evidence_ has shown to be true: yes,
there are costs to NAT, but it also has benefits, in that it attacks some of
the fundamental flaws in the IPvN architectures in general (lack of local
allocation of identifiers, ability to relocate [aka renumber] without local
reconfiguration, etc) and in IPv4 in particular (not enough address bits),
and when people look at the overall cost/benefit ratio, they prefer it to the
alternatives.

        Noel