So, one of my concerns with SIDR and it's current scope
is this, which I'd very much like some clue about...

Today RIRs allocate address space and AS numbers but
have NO controls over the routing system.  IRRs are used
for routing policy specification, and ideally, data in IRRs is
used to derive router configuration elements.

If SIDR is deployed in some manner, RIRs (and possibly
IANA) would have the ability to control what gets routed, and
that's a capability that's never existed before, and one that I
would suspect operators take considerable issue with.

For example, if operators subscribe to a SIDR model and an
RIR gets owned and an allocation revoked, or if they happen
to make a mistake that results in the same, or some bill non-
payment issue arises, or can't publish an update and one
expires, or whatever, then the result would be that the
associated routing for the concerned prefixes would break.
That's never been a threat vector before for ISPs.  They
wholly control and authorize what they announce.

What's worse is regarding who holds those keys.  If some
country holding the keys (TA) goes to war with another and
decides they want to revoke all of their allocations, then ISPs
would have zero control over this outside of their own routing
domain.

One might argue this is part of the reason why DNSSEC
raises deployment concerns for some, and has driven folks
like .cn to deploy their own root.

In my mind, a web of trust model for the routing system better
mirrors what we've got today.  Allocation authentication is
necessary, but just that.  IRRs and RPSL or even some back
end route server type functions down the road, else we'll
never get deployment of this.

So, I'm sure suspect I'm missing something here, could folks
please help me better understand both incremental deployment
models and how the above isn't an issue?

Thanks!

-danny
_______________________________________________
Sidr mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/sidr

Reply via email to