[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

What's really needed is a routing slot market

On Sun, Feb 6, 2011 at 11:15 AM, Joel Jaeggli <joelja at bogus.com> wrote:
> So assuming this operates on a pollution model the victims of routing
> table bloat are compensated by the routing table pollutors for the use
> of the slots which they have to carry. so I take the marginal cost of

In this case the "victims" are the other ASes,  and the "pollutors" are the
ASes that announce lots of blocks.  However, the pollution is also
something the
"victims" need,  so it's not really pollution at all,  unless an
excessive number
of slots are used  it's more  "meat"  than trash,  the pollution model doesn't
exactly make sense;  the basic announced routes for connectivity are
not like pollution.
They are more like fertilizer... nutrients that are absolutely
essential when utilized in
appropriate quantities, but harmful in excessive quantity.
And if too many use them in excessive quantity, then polluted runoff
is released as a side-effect.

There is an assumption that waste is so rampant, that a per-slot cost
would lead to efficiency,
and no loss of connectivity or stability,   but there appears to be
lack of data validating the suggestion.

Private "routing slot markets" could be a huge can of worms... and we
thought peering
spats were bad.  Some $BIG_DSL_PROVIDER  is going to refuse to pay some
$BIG_HOSTING_PROVIDER (or anyone else)  for their routing slots, they
will know that
their size makes it too unpallatable  for anyone else on the market to
_not give them_ the
slots,  even if they are one of the larger polluters with numerous
Other providers simply can't afford to be the provider whose customers
If   $BIG_HOSTING_PROVIDER's   routers do not have $BIG_DSL_PROVIDER's
$BIG_HOSTING_PROVIDER's  customers will scream, and jump ship   for
that has $big_dsl_provider routability.

There will be other $BIG_COMPANYs   that as well have superior
negotiating position, and
noone else will be in a position to discard their routes, when they
refuse to pay,  or negotiate a price
that reflects their superior position  (rather than one reflecting
cost of their excessive use of routing slots).

So first of all...  if there's a buying and selling of routing slots
a "market".
It cannot be a voluntary market,  or it will simply lead to  a chaotic
situation where
numerous big providers get free routes, and everyone else has to pay
the big providers
extortionate/disproportionately high fees because they _have to have
those slots_   due to so many of their
hosting customers requiring   $big_provider connectivity.

To ascertain a market in the first place... need to know....
How is the number of slots that will be on the market determined?

Who gets to initialize the market; create and sell  paid 'routing
slots',  and  what will give them the
power to enforce all users of routing slots buy from them...

Are these one-time purchases...  or do  'routing slot' purchases incur
maintenance fees?

How would the 'ownership' of a slot get verified, when a route is announced?

How is it decided how much cost 'repayment'  each AS gets?
Who is going to make sure each AS fully populates their table with
each routing slot they have been
paid to fill  and   they do not populate any slots that   were not
purchased by the originator on the open market?

The idea of 'ownership'  of other people's things  (slots on other
people's routers) generally requires the AS owning those things to
sell them. That would suggest each announcer of routes having to go to
each AS and paying each AS a fee for slots on their routers ---  not
only would it be expensive,  but the communication overhead required
would be massive.

So clearly any market would need to be centralized;  transactions
would need to happen through one entity.
One  buy/sell   transaction for  a routing slot     on  _all_
participating ASes.

Seems like a tall order

> the slots that I need subtract the royalities I recieve from the other
> participants and if I'm close to the mean number of slots per
> participant then it nets out to zero.