Observations of an Internet Middleman (Level3) (was: RIP

Matthew Petach mpetach at netflight.com
Mon May 19 04:15:38 UTC 2014

On Sun, May 18, 2014 at 11:40 AM, Owen DeLong <owen at delong.com> wrote:

> Traffic Symmetry is a distraction that the $ACCESS_PROVIDERS would like us
> to
> focus on.
> The reality is that $ACCESS_PROVIDERS want us to focus on that so that we
> don’t
> see what is really going on which is a battle to deeper (or avoid
> increasing peering
> capacity with) networks they think they can force to pay them more money.
> This is an age old tactic and it isn’t unique to $ACCESS_PROVIDERS. The
> larger
> $BACKBONE_PROVIDERS did it in the past, too. The first one was a railroad
> company turned telecom. Then came the remnants of PSI. Today, it’s the
> largest
> $ACCESS_PROVIDERS. Usually, this just results in harm to both sides and
> eventually a loss of subscribers. The $ACCESS_PROVIDERS have an advantage
> in the latter as they mostly avoid loss of subscribers through the fact
> that the
> subscribers don’t have anywhere else that they can usefully go.
> Owen
I agree it's a distraction from the primary reason
behind it; today, networks using traffic ratios as the
excuse why peering 'doesn't make sense'; even if
the traffic ratios are balanced, though, I'm sure there
would be another requirement, such as minimum
number of prefixes announced (mass deaggregation
should meet that one), minimum number of downstream
ASNs announced (a 4-byte ASN for every rack
switch cluster should handle that one), minimum
backbone size (isn't everyone already doing 100G
at this point?), present on multiple continents (isn't
everyone?).  When you get right down to it, though,
it's all just hand waving around the age-old question
of "how many entities can I push to pay, without going
too far, and alienating the entire rest of the internet?"
In years gone by, that line was relatively conservative;
hosting spammers, for example, was thought to be a
sure kiss of death that would lead all other networks
to shun you, effectively cutting you off from the
internet community.  However, I think we've all seen
that our notion of the power of the community was
overstated; internet-wide shunning didn't materialize,
we failed at being able to cut spammers off to a level
that would make it unprofitable, and we still have thread
after thread about BCP 38 compliance.  Seeing that,
it's really not surprising that networks would become
bolder, no longer fearing widespread disapproval or
community disaffection for actions that might have
seemed more extreme in years past.  I mean, at this
point we can't even seem to effectively shame people
into not leaking deaggregated prefixes, in spite of the
weekly email to the list naming names, and in spite of
Patrick's exhortations.  Given all that, it really isn't all
that suprising that a certain 3-digit ASN is trying to pull
games like this, refusing to augment capacity in the
hopes they can use their customer base as leverage.
They've realized the internet has no teeth, so they can
act with impunity.  It's sad to see, but somehow, it's not
all that surprising.

So yes, Owen--I agree with you; it's not a new tactic,
it's just being carried out more brazenly and with less
and less fear of community opprobrium.


More information about the NANOG mailing list