ietf
[Top] [All Lists]

Re: As if you don't have enough to read..

2015-03-13 12:11:52
John Leslie wrote:
Miles Fidelman <mfidelman(_at_)meetinghouse(_dot_)net> wrote:
John Leslie wrote:
Miles Fidelman <mfidelman(_at_)meetinghouse(_dot_)net> wrote:
Wait a minute... isn't the Internet "capital I" defined precisely by the
collection of IP addresses that are reachable from each other?
I'm not aware of any definition which says that...
It's kind of implicit.
    (Isn't "kind of implicit" the opposite of "a definition"?)

It's a mutually interconnected address space.
    No, it isn't...

I think we have to disagree on this.



Just like any telephone connected to the PSTN.
    It's not remotely similar to a telephone connected to the PSTN.

    It's connected to a network connected to another network connected
to yet another network (et cetera), none of which have any fixed
contractual interconnections. Paths through the network of networks
come and go (mostly) without any human intervention or even awareness.

Have you actually looked at the internals of the PSTN, particularly packetized underlying infrastructure, VoIP and all that? Any long distance phone call, particularly an international one, gets packetized, and goes through lots of different networks.

As to contractual relationships - what do you call backbone peering?

The analogy is very direct.


(Yes, NAT confuses things a bit - but arguably it's the public address
of a NAT device that's the "Internet endpoint").
That is a reasonable interpretation. But, I must disagree that there
is any generally-accepted definition of the term "Internet endpoint".
I've certainly seen the term used interchangeably with IP address in
multiple contexts - mostly around DDoS attack on a particular "internet
endpoint" or another, and I seem to recall a draft MIB for "internet
endpoints" that essentially treated the term interchangeably with IP
addresses,
    Can you give an example?

Google is your friend:
https://tools.ietf.org/html/draft-ops-endpoint-mib-00
and no less than Doug Comer uses the term in his classic "Internetworking w/ TCP/IP"
https://books.google.com/books?id=yhwfAQAAIAAJ&q=%22internet+endpoint%22+%22IP+address%22&dq=%22internet+endpoint%22+%22IP+address%22&hl=en&sa=X&ei=OhgDVaarH4qlNoKggfAH&ved=0CDUQ6AEwAQ
(though he uses it in the context of a specific socket)


or a physical port with an IP address. If NOT an IP address, is there
really any other viable interpretation?
    "Internet endpoint" generally refers to a node with an interface
connected to one of the networks of "the Internet". "IP address" is
a transitory characteristic of the interface. These concepts seem
different to me...

(And yes, it would be nice if the FCC order defined its terms
    It's not a question of "nice" -- the FCC _must_ define its terms.
Judges are not permitted to guess.

Hah.


- though at one point the order says this "?Fixed? broadband Internet
   access service refers to a broadband Internet access service that
   serves end users primarily at fixed endpoints using stationary
   equipment, such as the modem that connects an end user?s home router,
   computer, or other Internet access device to the network).
    Yeah, they screwed that definition up, too...

Please understand what ISPs actually do:

1. We receive packets and _try_ to route them to another node which we
    have reason to be "closer" to the destination address;

2. We advertise our "closeness" to particular ranges of IP addresses.

That's it, folks. Whatever else we do in support of these cannot
change the fact that we cannot deliver packets "to" most IP addresses;
and we cannot even know whether a packet we may deliver to a customer
is actually "from" a particular IP address (least of all whether it
is in response to a packet our customer asked us to forward).

It's all "best effort" -- which means we make no representation
whether any packet will reach the nominal destination.
One expects a bit more than that.
    Regardless of "expects", that's what we do.
One expects an ISP to actually have connectivity,
    I suppose, but "connectivity" typically coveres multiple paths with
differing characteristics.

peering agreements,
    Not all ISPs do "peering agreements" -- though probably most if not
all of the ISPs the FCC seems to intend to regulate under Title II
probably do. In any case, peering agreements are mostly secret, and
ISPs have good and valid reasons to keep them that way.

and routing tables
    Yes, we all have routing tables.

set up that let a customer exchange packets with addresses across
the Internet as a whole.
    No, we don't. Routing tables are one-way things: where do we forward
a packet hoping the node we forward to is closer? The routing tables
that affect a return packet contain wildly different data.

At least, I've never come across an ISP that advertises connectivity
only to a limited set of ASNs,
    That's a meaningless concept: no ISP has control over what happens
to a packet after it is forwarded to a different autonomous system.

or a limited set of IP addresses
    Again, no ISP can control what happens to a packet after it is
forwarded. ISPs _do_ filter traffic to particular IP ranges, but not
because they advertise that, rather to mitigate denial-of-service or
possibly inhibit spammers.

(well, there are enterprise networks and VPNs, but those are not
generally viewed as ISPs).
And things are generally considered broken when a polluted routing
table leads some part of the net to become unreachable.
    When such "brokenness" persists, most ISPs will "route around it."

And it's generally considered a "bad thing" that governments, like
the Chinese, put up big firewalls that block traffic to large chunks
of the net.
    Obviously, that is beyond the control of ISPs -- but it's really
a tiny slice of what is beyond their control...

The same argument applies to phone service.  I may get my local phone
service from Verizon, but I expect to be able to place a call to any
phone, anywhere in the world, that's connected to the PSTN. Verizon is
providing the "capability" to dial and reach all those numbers, but it's
not providing the end-to-end connectivity.  If they started saying "you
can't call numbers in Cincinatti, or Italy" I think we'd all agree that
something was wrong.
    No comment -- different topic entirely.

Disagree completely. Package delivery is not phone service, but they are both common carriers; and now so are ISPs. Not a different topic at all.



====
    I understand you _want_ Internet Service to fit the FCC definition.
But wishing doesn't make it so. The FCC definition here calls for things
that ISPs are completely unable to do.



By your reading.  Not by mine.

Miles Fidelman






--
In theory, there is no difference between theory and practice.
In practice, there is.   .... Yogi Berra