Re: [rrg] Last call for revised ILNP documents

[Steven Blake <[e-mail]>]

On Sun, 2012-01-22 at 21:03 +0100, Stephane Bortzmeyer wrote:

> On Sat, Jan 14, 2012 at 10:18:03AM -0800,
>  Tony Li <[e-mail]> wrote 
>  a message of 27 lines which said:
> 
> > Given the extent of the changes, it's appropriate that the RG review
> > and comment, with a full two week review cycle and consensus check.
> 
> I've read the documents (except the IPv4-specific ones) and I believe
> they are mostly complete and clearly explain how ILNP works.
> 
> There is just one thing which seems to be absent: I do not find
> clearly stated how are the identifier and locator stored in the IPv6
> packet. From some paragraphs, I guess the locator use the high-order
> 64 bits of the IPv6 address and the identifier the low-order bits but
> I was not able to find it clearly written.

It appears that some text from draft-rja-ilnp-intro didn't make it into
the reorganized drafts.  I suggest the authors write a very short draft
describing the ILNPv6 header.

> Another point is more a detail: the documents seem to mandate that
> there is a state in the machine for every ILNP correspondent (the
> ILCC). This is no problem for a HTTP server (there is already the TCP
> state) but more problematic for active authoritative DNS servers:
> today, they are able to sustain a very high activity because they have
> absolutely zero state. May be the ILCC could be made optional for
> stateless servers? (There is no need to receive locator updates for
> the ultra-short transactions of the DNS.)
> 
> Otherwise, there is a typo in the references,
> draft-irtf-rrg-ilnp-noncev6 is written draft-irtf-rrg-ilnp-nonce6 in
> the docs (no 'v').

It makes no sense to use the ILNP machinery for a DNS query (or any
similar UDP-based simple request/response transaction).  One of the
drafts probably ought to state this explicitly.

[Yes, I know I am tardy submitting comments.]


Regards,

// Steve