Short version: Ran doesn't like debate - and blames me for the
debates which occur on the RRG. Apparently this
debate-ridden Research Group is a new experience
for him.
He expects people to understand ILNP by, in part
reading academic papers - such as one which for
many people is behind a paywall and which he
doesn't mention on the RRG list, and only references
once in a non-specific manner in one of his I-Ds.
He wants people to understand and accept his
ideas - but I don't recall him discussing,
critiquing or debating anyone else's architecture.
ILNP for IPv4 involves some very different
practical considerations for ILNP for IPv6 -
including the need to upgrade some or all DFZ and
other routers. It also imposes a 28 byte overhead
on all packets.
As far as I know, Ran has never argued why the
extra host burdens, extra packets, longer packets
(ILNP for IPv4) and typically extra delays in
establishing communications for all hosts are
justified by avoiding the extra network
infrastructure of alternatives such as LISP, Ivip
and IRON-RANGER.
Nor has he compared ILNP for IPv4 with the somewhat
similar architecture hIPv4.
Hi Ran,
You wrote:
> (The title is incorrect. ILNP for IPv4 uses an IPv4 option.)
Fixed.
You wrote (msg06540):
Unlike the IETF, the IRTF expects that participants already know
the RG topic very well, and that participants WILL take the time
to read the relevant research literature.
So you will be aware of Patrick Frejborg's recent work on hIPv4.
This is not HIP. (I mistakenly used upper-case H in previous
messages.)
How does ILNP for IPv4 compare with hIPv4? Both involve using new
IPv4 header options to convey extra "Identifier" or extended address
bits to hosts with modified stacks.
>> This approach would be significantly different from
>> the IPv6 approach. So whatever claims are made
>> about ILNP in IPv6 don't necessarily hold for its
>> application to IPv4.
>
> The *architecture* is IDENTICAL. The only thing that
> changed is where the bits get stuffed on the wire.
For a sufficiently rarefied definition of the boundary between
"architecture" and everything else, yes, the architecture is identical.
But relying on IPv4 header options has important practical consequences.
You acknowledge some consequences - the need for modified ARP in the
local network of all hosts. You also acknowledge that routers -
including DFZ routers - may handle packets with IP options in the
"slow path", which is entirely prohibitive.
So it seems that to implement ILNP for IPv4 at all, there would have
to be upgrades to some or all DFZ routers, just to ensure that they
handled the packets with your new IP option via their normal FIB
paths. You also need to upgrade the stacks of all hosts. That's two
huge sets of obstacles to adoption.
ILNP for IPv4 would add a 28 byte overhead to every packet. This
adds overhead in general and also reduces the maximum payload a
packet can handle. The lack of such overheads or restrictions is one
of the good things about ILNP for IPv6.
As with ILNP for IPv6 and with other CEE (Locator / Identifier
Separation) architectures IPv6, hosts need to do more work than they
do today, including doing mapping lookups, which involves extra
packets and delays in establishing initial communications.
Where do you argue that this burden, for all hosts, is a worthwhile
price to pay for avoiding the additional network infrastructure such
as is required by Ivip, LISP or IRON-RANGER (CES architectures)?
These architectures involve no host alterations - and do not require
changes to all networks.
>> However, if ILNP is to work with ordinary
>> applications, then it seems that it is stuck with
>> a 32 bit limitation on Identifier. (As best I can
>> guess how it might work.)
> Not true.
OK - now I read this part of the paper, I can see you keep the same
64 bit Identifier.
>> Please email me a copy of the paper.
>
> Unfortunately, Copyright Law (in Australia and elsewhere) means
> that I can't legally do so.
OK. However, you have known since before October last year how you
are contemplating implementing ILNP for IPv4. In all that time, you
have failed to explain this on the list. All you had to do is quote
the text I quoted, or write something similar - on the RRG list
and/or in one of the ILNP I-Ds.
>> To use this, hosts need new stacks. Are you planning on this working
>> with existing applications?
>
> Yes. No application changes will be needed.
OK, as with ILNP for IPv6.
>> You can't expect me or anyone else to read all the papers
>> listed at your site:
>>
>> http://ilnp.cs.st-andrews.ac.uk/
>>
>> some of which go back to 2004, just to understand your proposal.
>
> I precisely do expect that.
OK. I stand corrected. You do expect this.
I think your expectation is entirely unreasonable - especially since
at least one of these papers is behind a paywall (USD$34) for those
people who don't have company or university access to the journal.
>> It is reasonable to expect anyone interested in your proposal to read
>> the Internet Drafts, summaries etc. you post to the list. People
>> can't be expected to keep up with every update to these, but I did
>> read your I-Ds at the start of the year and made a concerted attempt
>> to understand your proposal then.
>
> You apparently didn't read draft-rja-ilnp-dns,
> which describes the L32 record used for ILNPv4.
I read version 01 it in January when I spent most time discussing
ILNP. That doesn't mention IPv4. Version 02 (2010-02-08) does
mention this 32 bit Locator field. However, there's no mention of
the packet format - so no way of understanding how ILNP would work
for IPv4 - unless of course I had happened to spend $34 on the
October 2009 paper, which I don't recall you mentioning on the list
until just now.
>> Please make the above mentioned paper available on your site, or
>> email it to anyone on the list who wants to read it.
>
> You are suggesting that I break the law.
You are expecting people to spend good money, as well as their
precious time, reading your material.
I am suggesting that you should make all the material any RRG person
needs to understand ILNP freely available. If that paper is tied up
with copyright, then I suggest you write a separate document as an
I-D, a mailing list message or a document on your site:
http://ilnp.cs.st-andrews.ac.uk/
>> Sure - no-one is obligated to do so. Its just that if you want your
>> proposal to be widely understood and respected, as far as I know,
>> there is no alternative to patient and detailed discussion and debate.
>
> I believe a majority of list understands ILNP well enough already.
I think you are mistaken, but there's no way of proving either
position. How many of them understood how you proposed to implement
ILNP for IPv4?
> Certainly the off-list queries about very minor specific details
> indicate that the folks sending those queries already understand
> ILNP well. People attending RG meetings tell me they understand
> ILNP (whether or not they prefer it).
How can you be sure they understand it? They may understand some of
what you wrote, or think they understand it - but actually not
understand it properly at all. How do you know what they read? They
may have read a subset of your material, but not something important
like the section I quoted - section 5.2 on pages 14 and 15 of a paper
which is listed on your site, and only referenced once, in an
entirely non-specific manner, in draft-rja-ilnp-intro-03.
> People in Anaheim were not
> necessarily enthusiastic about the Co-Chair's decision, but there
> were no complaints voiced about ILNP documentation or clarity in
> Anaheim
OK, but you went to a lot of effort on the list to complain about my
lack of understanding, while doing nothing to help me understand
something important about ILNP. At any time since December, you
could have posted text such as I just quoted from your paper. In the
end, all you did was loosely refer to a paper - and expect me to
chase it up.
> You appear to be a statistical outlier in this regard.
Maybe I am fussier about my understanding than some other folks.
> Probably
> not the only person, but certainly you aren't in the majority.
It would be better if more RRG people went to the trouble I did to
understand and comment on all the proposals, including yours. It
would be better if you critiqued other proposals.
>>> RGs are not setup as debating forums.
>>
>> Yet they are forums in which debate frequently does occur.
>
> This one sadly seems to have become one, thanks to your presence.
> Other RGs I've been involved with haven't been debating societies.
I have never asserted the RRG is a society. I do assert that it is a
forum for debate - and furthermore that this is essential to the
RRG's work.
So you don't like debate - and indeed you blame me for the
preponderance of debate which occurs on the RRG!
As far as I know, "debate" is the same thing as meaningful discussion
when there are more than one point of view.
You seem to want people to understand and accept your ideas without
debate or meaningful discussion. I have never heard of such a
suggestion in any professional setting.
The Charter:
http://www.irtf.org/charter?gtype=rg&group=rrg
includes:
The RRG will have an open general discussion mailing list where
any topic of interest to the routing research community can be
discussed, and topics related to scalable routing architectures
are particularly encouraged.
It seems you are intent on discussion without debate. I regard that
as meaningless.
Furthermore, you only seem interested in discussing ILNP - though you
did make some constructive suggestions recently about what the RRG
might be able to achieve consensus on.
You want and expect other people to understand and ideally agree with
your ideas - other people who have their own proposals. Yet you
don't bother to critique anyone else's proposal.
Can you identify any other RRG participants who share your preference
for discussion without debate? Its my impression that most folks who
write to the list are keen to debate things.
- Robin
_______________________________________________
rrg mailing list
[email protected]
http://www.irtf.org/mailman/listinfo/rrg
