Well, while Sean's wording at times is, ahem, suboptimal, he is
referring to a real problem in the Internet System, and in fact one
that the North American Network Operators Group should take
responsibility for, at least for the North American part. The problem
is what others believe is a strength: an anarchic system of many
autonomous service providers with little or no service model/metrics
(neither for the local nor for the system level), and no well defined
rules of conduct for interoperation and problem resolution.
A few weeks back I had problems with getting to another CONUS site.
Not exactly capillary connectivity, source was at SDSC, destination at
FIX-West. My service provider is CERFnet. Some days after emailing
about the problem (it happened on a Friday evening that I had very
intermittent connectivity, with packets from my traceroute getting lost
in Sprintlink land) I got actually quite polite and professional
responses that, based on my traceroute, it was outside of CERFnet's
domain and best they can do is inform Sprint. As I network engineer I
understand the context and the issues. As a network user I do not find
the service model acceptable. I thought my service provider was selling
me Internet connectivity, not regional connectivity. This is not to
single out CERFnet or Sprint, given the way The System is run, the
service providers likely are all the same, at least above the 90th
percentile. Certainly because there is no global (or even domestic)
problem resolution procedure that has any leverage with other service
providers. Eventually something hits the roof for someone, and he
starts bitching more publicly as his only remaining leverage, which is
what seemed happening to Sean. What choice did he have? Run to the
government and ask for regulation, as the service providers just can't
get their act together at a systemic level? Well, that may not be one
of *his* first choices, but sooner or later Real Users (rather than us
networking bigots) get pissed. You know, those you sell services to?
That you give, say, T1 connectivity to without being clear what it
means to a user, who's traffic may have to traverse five ISPs? And who
may not be interested in hearing about finger pointing from the service
provider or deferred problem resolution? I mean, what would you say if
your electrical power would drop to 80 Volts (more than once in a year)
and your local power company would tell you "uh, that's not our fault,
it is just power company 2000 miles away ran out of oil, or misplaced
some nuclear refuelment rod?"
This all should be critical to the NANOG agenda if the service
providers would want to continue to provide quality services.
I think it is an attitude problem of people largely caring about their
own swamp.
IMHO, given that this *IS* an anarchic system, the proper place to get
pissed is in public, i.e. this list or another list. That is how anarchic
systems work! Because everyone is autonomous and makes their own
decisions, it is neccessary that the information required to make those
decisions be fully public.
In this case, new people are continually joining this anarchic system and
the only way they can understand it is to see it in action *IN PUBLIC*.
I'm not just talking about ISP's who are moving to multihoming and BGP,
but also new employees at existing NSP's.
Another characteristic of anarchy is that if you are smart and you know
something, then you get to impart that knowledge however you please and
there is no big brother out there who is going to tell you that this is
conduct unbecoming to a teacher. Well, Sean is imparting knowledge....
What I mean is constructivity. Repeated shows of faults does not
contribute to the solution. This is just whining in public.
So, here is my try to discuss what we all can do:
I propose:
Whoever has dial up customers, and cannot, for one reason or the other
convince them, to keep the interface up, should do the following:
1)- explain the customer that the link cannot go up and down all the
time.
- that rapid changes in the link will not be propagated, but result in
unreachability of his computer/network.
- that such an arrangement is not suitable to announce reachability of
his host to the outside, since nobody knows when he will actually be
reachable (so, no email directly to his address, don't advertise it as
ftp address, etc.)
2) pull the route up via a loopback interface and a bigger metric. When
the dial interface comes up, the route via the loopback will be
overridden by the better 'connected' route. this keeps the route flap
within the same router
This is not 'announcing lies', the customer 'lives' in that router. This
is a contribution to the stability of the routing environment.
I myself do not attach LANs via dial-up, and if, the route is aggregated
to the overall routes of the communication server where the customer
dials into.
The note that was forwarded was simply unbelievable at this day and
age. The Internet community has to pay a large price when a major
provider has a policy of not aggregating routes and also not using
statics for sole dialup customers. Such behavior should be made
readily apparent to perspective customers, resellers, and referrals
so that they might understand the technical capabilities and
marketing tactics of a given provider.
Sean is fighting the good fight to get people to do the right thing.
That he does so with gorilla tactics at times is probably
appropriate given in some cases the lack of response and in other
cases completely inappropriate response from various providers and
customers. As his customer I *very* much appreciate and encourage
his efforts.