Re: [lisp] I-D Action: draft-ietf-lisp-ddt-05.txt

[Joel M. Halpern]

I have a problem.  I was hoping to stay out of this.

But the proposed description does not work.
Since xTRs form overlapping sets, I don't see any way to assign IDs to 
Instances such that any given xTR does not need see the upper 8 bits of 
the instance-ID it receives.

If the argument were that we want future flexibility, so we are 
supporting 32 bits in the control plane, I could see it.  But as 
written, I would not know what to do in an Etr or ITR if I had an 
Instance ID with the upper bits set?  Do I truncate on send, and compare 
only the lower 24 bits upon receipt?  Once we start needing the upper 
byte, I don't see how that could work very long.

Yours,
Joel

On 5/2/16 1:10 PM, Dino Farinacci wrote:

On May 2, 2016, at 1:36 AM, Luigi Iannone <g...@gigix.net> wrote:


On 29 Apr 2016, at 23:18, Dino Farinacci <farina...@gmail.com> wrote:

So how about this as a proposal. The DDT doc refers to 32-bits for the 
control-plane and RFC 6830 refers to 24-bits in the data-plane and the LCAF 
document explains how to map the 32 into 24 and what the benefit is?

So I’m asking and offering to update the LCAF document to reflect this. Agree?


Sounds good to me. I just would like to see a ref to LCAF in the DDT document, 
so that we have a pointer toward were the 32 to 24 bits conversion is described.

How about this change? Please ack DDT authors and then supply a reference to 
draft-ietf-lisp-lcaf-13.

Once I get an ack, I’ll submit this post-last-call draft update.

Dino











ciao

L.


Dino

The change was consequence of the following comment that I made to the authors 
as shepherd:

Instance ID:
————————

This document define the IID as a 32 bit field, how does it work with the 24 
bit IID defined in RFC6830?
I would suggest that either you provide discussion on the above point or you 
refine the IID as 24 bit.


The main point being: how do we shrink the 32bits IID in the 24bits field of 
the LISP header?

As for the explanation of Dino in this thread, the best thing is to leave 32 as 
value, but I would like to see an explicit explanation on how to go from 32bits 
 to 24 bits, referencing the LCAF document. It is just a clarification sentence 
to add.

ciao

L.







On 27 Apr 2016, at 22:47, jmh.direct <jmh.dir...@joelhalpern.com> wrote:

Authors, was there a working group request for, or review of, this change?

Yours,
Joel



Sent via the Samsung Galaxy S® 6, an AT&T 4G LTE smartphone
-------- Original message --------
From: Dino Farinacci <farina...@gmail.com>
Date: 4/27/2016 4:44 PM (GMT-05:00)
To: "Joel M. Halpern" <j...@joelhalpern.com>
Cc: Anton Smirnov <asmir...@cisco.com>, lisp@ietf.org
Subject: Re: [lisp] I-D Action: draft-ietf-lisp-ddt-05.txt

On Apr 27, 2016, at 1:36 PM, Joel M. Halpern <j...@joelhalpern.com> wrote:

I am a bit confused.  I suspect other working group members are as well.
The DDT document completed WG last call some time ago, and was waiting for some 
final edits, which were I believe just done.
The LCAF document has completed last call.

Well I lost track of the DDT document status.

Dino, which document are you requesting be modified?  What modification are you 
asking for?

I am requesting a single change (in 2 places, see below). A change back from 
24-bits to 32-bits describing the instance-ID. I don’t know why the change was 
done during this late stage in the draft. To me, that is a huge change.

Or have I got it backwards, and Anton is asking for a modification?  As I said, 
I got lost.

I reviewed the changes in -05 and noticed this:






And the change should not have happened since our intention at the very 
begninning was to have 2**32 VPNs.

There was no justification for this change and it happened very late in the 
process.

Dino


Yours,
Joel

On 4/27/16 4:25 PM, Dino Farinacci wrote:
Can you make the change so we can try to advance the document to last call?

Dino

On Apr 27, 2016, at 1:23 PM, Anton Smirnov <asmir...@cisco.com> wrote:

we will consider this input for the next doc revision.

Anton


On 04/27/2016 07:06 PM, Dino Farinacci wrote:

Hi Dino,
since XEID prefix is not seen anywhere on the wire these words should not be 
viewed as normative; more like guidance for implementers. For DDT specification 
itself it is not important if IID is 24-bit, 32-bit or any other bit length. 
DDT relies on other control plane specifications (notably LCAF draft) to 
specify how IID looks like and how it is propagated in control messages.

If that is the case, why is the length included in the text then? I disagree 
though, the length is critically important because it conveys the maximum 
number of VPNs, per mapping system, that can be supported.

LCAF draft currently depicts 32-bit space to store IID on the figure but then 
goes on saying:

Instance ID:  the low-order 24-bits that can go into a LISP data
  header when the I-bit is set.  See [RFC6830] for details.

Right, because that is the only way to fit 32-bits into 24. ;-)

So IMO the ambiguity comes from the LCAF document. draft-ietf-lisp-lcaf should 
be more specific on IID length. Furthermore, if LCAF draft explicitly defined 
IID to be 32-bits then it should discuss what to do with excess bits in case of 
LISP encapsulation.

No, this is not true. And you might not have the history of DDT. But we put 
32-bits in the DDT document and then had the encoding in the LCAF document 
reflect that.

If DDT draft progresses before LCAF draft then it is more correct to be 
compatible with existing RFCs in saying that IID is a 24-bit value. DDT doc 
does not look like a proper place to redefine IID length from 24-bit to 32-bits.

The LCAF draft just ended last call and is going to IESG.

If you strongly disagree with above then to unblock DDT spec from LCAF ambiguity we 
may remove explicit mention of IID bit length from DDT spec and put something like 
"IID as defined by the LCAF draft”.

You can’t remove it. You have to make it 32-bits otherwise you created an 
inconsistency that is (1) not needed and (2) for no good reason.

I suggest you leave that text alone and keep it at 32-bits.

Dino


Anton


On 04/25/2016 09:49 PM, Dino Farinacci wrote:
Authors, this change:


Is actually incorrect (change change is from 32 to 24). We have 32-bit
Instance-ID encodings in the LCAF Instance ID Type and want to support
that length in the control-plane EVEN THOUGH the data-plane can only
hold 24-bits.

Meaning, if you use different mapping systems, you can actually reuse
instance-IDs. This reuse was part of our initial intention.

Dino


_______________________________________________
lisp mailing list
lisp@ietf.org
https://www.ietf.org/mailman/listinfo/lisp


_______________________________________________
lisp mailing list
lisp@ietf.org
https://www.ietf.org/mailman/listinfo/lisp





_______________________________________________
lisp mailing list
lisp@ietf.org
https://www.ietf.org/mailman/listinfo/lisp