Re: [RAM] Re: the separation of ID/RLOC

[Scott W Brim <swb at employees.org>]

On 07/03/2007 05:34 AM, Eliot Lear allegedly wrote:
> I think a reasonable question to ask is at what layer mobility should be
> provided.  When mobility is provided in  HIP, for instance, one needs to
> handle caching concerns within DNS and of course have appropriate
> bonafides available to authorize an appropriate record update.  In
> effect, a DNS server becomes the initial home agent, and that
> responsibility is subsequently transfered to the mobile node as handoffs
> are made.  This of course assumes that two devices don't move at the
> same time but that seems quite reasonable.

You've just pointed out the problems with making DNS the HA.  DNS is
for resolving name to identifier.  HAs are for resolving identifier to
current locator.  The authorization needed to make a DNS change should
be different in scope from the one required to change state in a HA.
It's good to have the two different phases of the process somewhat
decoupled.

> If we place mobility in LISP, the caching semantics of LISP-CONS would
> need to support it and whatever mapping binding would of course need to
> be secured, or an update would need to be made to NERD (NERD simply
> doesn't do mobility, quite frankly).

The caching that needs to be done in LISP for (fast) mobility) is
luckily just in the ITRs.

See you ... Scott

_______________________________________________
RAM mailing list
RAM at iab.org
https://www1.ietf.org/mailman/listinfo/ram