Re: [lisp] Fwd: Comments about draft-cheng-lisp-shdht-01

[[e-mail] (Noel Chiappa)]

    > From: Michael Hoefling <[e-mail]>

    > I assume that the proxy reply mode should not be the default operation
    > mode.
    > ... no P-bit not proxy.

That's my understanding - but sometimes I get a little confused on these fine
details! So if I have made a mistake, I trust someone will let us know! :-)

The whole question of having to get the mappings from the ETRs is one I have
thought about a fair amount, because it (currently) makes it impossible to
cache mappings in, e.g., the MR.

(Let me quickly clarify that "currently" - although I have thought about how
to cache mappings, I don't believe there are any plans to do so at the moment
- at least, not that I am aware of. It would require changing the Map-Reply
format, to indicate which ranges of 'source' EIDs a particular mapping is
valid for. So unless we have to change the Map-Reply format for some other
reason - e.g. to include signatures on mappings - my guess is that this is
unlikely to happen. And it's not clear if we need to cache mappings at
intermediate points between the authoritative source [the ETR] and the user
[the ITR] - although the DNS does have this capability, IIRC. Well, that's
the kind of thing we will discover through trying it...)

It is a slightly odd design choice in a way - it's a bit of an odd form for
the subsystem boundary (compared to other systems which provide mappings). It
wasn't until I started thinking of the ALT/etc as 'indexing systems' (which
is not how they were orginally described, I think) that it started to be
easier for me to think about. 


    >> This is necessary for things like 'source-specific mappings' (where an
    >> ETR returns different mappings to different ITRs).

    > Isn't this possible in proxy mode as well or is the selection logic
    > specific for ETRs?

Well, there is no way in the ETR/MS interface to specify that there are
multiple mappings, and which mappings should be given to which sources. So,
it is necessarily onl the ETRs which can provide source-specific mappings.

    > How is this currently achieved in DNS?

Do you mean 'how does DNS return different mappings to differerent lookup
requests' (as I gather a few people do, for load-balancing reasons, etc etc).
I don't know the details - I assume it's implementation-specific.

	Noel