[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [MEXT] multiple coa draft 11



Hi Hesham/George,
While I agree with you that this may not be useful in all scenarios, I think that SHOULD NOT implies that there is something inherently harmful with doing this. I simply fail to see any harm in allowing this. I was happy with the old text that said nothing :-)

Thanks
Suresh

On 12/03/09 06:08 AM, Hesham Soliman wrote:
FWIW I think SHOULD NOT strikes the right balance too, so count me in.


On 12/03/09 8:53 PM, "George Tsirtsis" <tsirtsis at googlemail.com> wrote:

I think I still agree with Hesham on this.

What Gerardo proposed and Ben expanded on, seems to provide at best an
optimization in terms of message length, with no difference in either
the number of exchanges or the functionality achieved.

BIDs, as well as FIDs, are just sort-labels identifying CoAs and Flow
Descriptors correspondingly.
I think we all agree that it is "possible" to associate multiple BIDs
with the same CoA. With the same way of thinking though it is also
"possible" to associated multiple FIDs with the same Flow Descriptor.
Should we do that too? I do not think so. Should we absolutely forbid
it? I do not think so either.

As I have stated before I think most of this does result necessarily
in incompatibility. It is mostly about emphasis and what the design
intention is. I think the drafts should make very clear how the draft
is intended to be used in the simple more generic case: i.e.,
- identify each available CoA with a BID --> one to one CoA-BID
mapping is a SHOULD
- identify each Flow of interest with a FID --> one to one
FlowDesc-FID mapping is a SHOULD
- Point FIDs to BIDs as needed.

This basic capability is already VERY advanced if you consider the
state of mobile Internet deployed today. So, limiting the obvious ways
of using this (using SHOULDs), while not forbidding other less-obvious
ways of using this, might not be a bad idea for now.

Regards
George

On Thu, Mar 12, 2009 at 7:45 AM, Hesham Soliman
<hesham at elevatemobile.com> wrote:

[Ben] You are right to say that in MCoA bulk registration is not
supported to CN.  But, MN can always based on its BCE to determine if it
should send a bulk BU or individual BUs to the destination.  If the
destination is CN for the above case, MN would send 2 individual BUs to
change CoA3 to CoA9.

I agree that there is a tradeoff for such optimization but my point is
that we should not disallow binding a CoA to multiple BIDs in the MCoA
draft.
=> I thought the discussion was circulating around one valid scenario for
not disallowing this. But in this scenario there doesn't seem to be much
benefit anyway. That's why I was questioning the need to worry about this.

Hesham


Regards,
Benjamin Lim


Hesham

Hence, flows for FID2 and FID3 arriving at HA will now go to CoA3.

Regards,
Benjamin Lim

Hesham Soliman wrote:
On 12/03/09 12:58 PM, "Benjamin Lim" <benjamin.limck at sg.panasonic.com>
wrote:

Hi Hesham,

=> I think this is a misunderstanding. This can be done with 1 message
anyway. Here it goes:

FID1 -> BID1->CoA1
FID2, FID5, FID7 -> BID2->CoA2
FID3 -> BID3->CoA3

Now CoA2 disappears or is no longer needed ...etc. The MN sends a BU
that includes the following:
CoA3, BID3, FID2, FID5, FID7, FID3
Which results in the following binding:
FID1 -> BID1->CoA1
FID2,FID3, FID5, FID7 -> BID3-> CoA3
The example you give above is for possibility 1 in Gerardo's mail.  From
what I understand in possibility 2,  MN need not send a BU with FIDs.
All MN send is a normal BU that indicates change of CoA.  So the BU
(after CoA2 cannot be used) would be something like BID2->CoA1.
=> BID2 -> CoA ? That doesn't map to the BCE you have below.

Hence, in the BCE at HA, it becomes:

FID1 -> BID1->CoA1
FID2, FID5, FID7 -> BID2->CoA1
FID3 -> BID3->CoA3

Do you feel that this would cause confusion at HA?
=> Not only will the HA be confused, I'm confused! How do you think this
will work? Can you send me step by step messages and be clear on how the
HA
handles having multiple FIDs pointing to multiple BIDs at the same time?
Until we see this, I don't think this scenario works at all unless I'm
missing something.

Hesham




Regards,
Benjamin Lim

Hesham Soliman wrote:
Thanks, my response inline:


On 12/03/09 12:18 PM, "Benjamin Lim" <benjamin.limck at sg.panasonic.com>
wrote:

Hi Hesham,

Mail was sent by Gerardo last week (friday on my client).  I snip his
respond.

<snip>
As Christian mentioned there has been some discussions about the
opportunity of having same CoAs for multiple BIDs. This may have a
some advantages in some specific flow binding scenarios. However this
is really something related to flow binding and I am not sure how to
address it here. I think a MAY is fine as it does not create any
interoperability issue.

As a background, the scenario is the following:
- MN has three interfaces with BID1, BID2, BID3 with some FIDs
associated to each BID
- Let's assume the interface (e.g. WiFi) for BID2 loses coverage. In
this case you have two possibilities:
1. the MN deregisters the BID2 and maps one by one all FIDs to other
BIDs
=> There is "no one by one" mapping, they're all done at the same time.

2. the MN binds the CoA of BID1 also to BID2 in order to avoid
sending
all the FID options. Automatically the FIDs are still bound to BID2
but with an active CoA
=> I think this can cause a lot of confusion, but luckily we don't need
to
go there. Please see below. You're assuming that the HA somehow knows
where
to map FIDs when they're bound to multiple BIDs. I don't understand how
you
think this will work. Perhaps you can send a step by step explanation.

In some scenarios option 2 gives you an optimization in terms of
signaling.
=> I think this is a misunderstanding. This can be done with 1 message
anyway. Here it goes:

FID1 -> BID1->CoA1
FID2, FID5, FID7 -> BID2->CoA2
FID3 -> BID3->CoA3

Now CoA2 disappears or is no longer needed ...etc. The MN sends a BU
that
includes the following:
CoA3, BID3, FID2, FID5, FID7, FID3
Which results in the following binding:
FID1 -> BID1->CoA1
FID2,FID3, FID5, FID7 -> BID3-> CoA3

Hesham



Gerardo
</snip>

Regards,
Benjamin Lim

Hesham Soliman wrote:
On 12/03/09 11:52 AM, "Benjamin Lim"
<benjamin.limck at sg.panasonic.com>
wrote:

Hi all,

<snip from Christian's mail>
I understand the SHOULD NOT to mean to the following:

1) The motivation for SHOULD NOT:  No known use case.
The above is not true.  We know of a use case with flow binding for
multiple BIDs to point to same CoA as what Gerardo describe.
=> Can someone point me to this case? I didn't see that email.
Gerardo?

Hesham

If we want to put text in the MCoA draft for clarity, I prefer to
say
something along the line that multiple BIDs MAY be assigned to the
same
CoA.

Regards,
Benjamin Lim

Suresh Krishnan wrote:
Hi George,

On 10/03/09 03:32 PM, George Tsirtsis wrote:
Hey Suresh,

Beyond calling this whole thing "useless" :-)

Maybe you want to also tell us why you want this.
I believe that the use case brought up by Gerardo is valid and this
is
one of the ways how I was intending to use the BIDs.

Also, tell us what you think about the downside discussed in this
thread i.e., having two ways of pointing flows to CoAs with both
FIDs
and BIDs .....as opposed with just one (FIDs).
I do not see this as a problem at all. As long as there are no
interoperability issues, I feel that we should not worry about
this.
In
other words, I fail to see the downside unless someone comes up
with
a
scenario/use case that will cause an issue.

Cheers
Suresh

_______________________________________________
MEXT mailing list
MEXT at ietf.org
https://www.ietf.org/mailman/listinfo/mext