Re: [Ideas] Mapping System Requirements and draft-padma-ideas-problem-statement-00.txt

[Dino Farinacci <[e-mail]>]

>> If registrations and requests are encrypted, then anyone could run the roots and the what goes in and out of the mapping system stays private. But there needs to be competition so the level of service stays at a high-quality production level.
> What is your vision? How much of the mapping data can be encrypted and
> how much information about the mapping owner can be hidden from the

Well I think we can encrypt the transport of messages from LISP sites to the mapping system. As to encrypting the stored state at the map-servers could also be done but here are some caveats:

(1) If the MSP is providing proxy-reply services it has to return Map-Replies to ITRs/PITRs. It can do so with lisp-sec for security. But the information needs to be stored in plaintext.

(2) All the map-servers need to know when they are not proxy-replying is to know the RLOCs of the ETRs of the site that registered the information (and not so much all of the RLOC-records that were registered) so the map-servers can forward Map-Requests to the ETRs so they can Map-Reply.

> mapping system operator? The ID cannot be encrypted as it is used as
> retrieval key. When we want to make sure that only rightful owners of

Right. At a minimum, the amount of plaintext that is stored in the map-servers are EID-prefix and the RLOCs in the RLOC-set (for case (2) above).

> IDs can register, the mapping system provider needs to authenticate the

That is done today with a Map-Register that contains an authentication hash across the entire Map-Register message.

> mapping owner. Can you elaborate the problem you are tackling and the
> solution in more detail?

I was solely asking if the messaging to the mapping system should be confidential.

Dino