Re: [MEXT] [WGLC] draft-ietf-mext-flow-binding-03

["Yuri Ismailov" <yuri.ismailov at ericsson.com>]

Hi George, all

George, what I understood from
*******
GT> I am not yet sure what non-trivial clarification we can provide.
IMO we do not need to provide any further clarifications on this subject.
********
is that you do not see any point of further discussion.
Well, I will anyway provide the text which I think is suitable for clarification.
On the other hand I disagree about the fact that TCP connection suffering performance is more clearer than TCP connection stop working. Strange that this is not obvious.

Below is the proposal for the text to add in the paragraph describing "Discard" action in chapter 4.2

After the text:

"1 Discard.  This value indicates a request to discard all packets
 in the flow described by the option.  No BIDs are associated with
 this Action."

Add the following text:

"In the case there is no correspondent FID to a particular flow or range of flows, which should be discarded, a mobile node MUST install a correspondent FID, otherwise error code 136 "FID not found" will be returned by HA. Care should be taken when issuing request for discarding packets as this will lead to disrupting applications communication after the "Discard" action was applied to selected flows. Implementations may consider notifying impacted applications in mobile nodes"



Regards
Yuri


-----Original Message-----
From: George Tsirtsis [mailto:tsirtsis at googlemail.com] 
Sent: Friday, October 23, 2009 5:54 PM
To: Yuri Ismailov
Cc: Laganier, Julien; mext at ietf.org
Subject: Re: [MEXT] [WGLC] draft-ietf-mext-flow-binding-03

Hi Yuri, comments inline.

On Fri, Oct 23, 2009 at 1:20 PM, Yuri Ismailov <yuri.ismailov at ericsson.com> wrote:
> Hi Goerge,
> Thanks for the answers,
> What I did not understand if there will be some clarification text in the draft around "discard" action or all that is obvious and does not require any changes? My comment was that some additional text is needed.
>

GT> I am not yet sure what non-trivial clarification we can provide.
IMO we do not need to provide any further clarifications on this subject.

> I do not think your clarification about using FID-PRI is quite clear. The reason is that there is no FID for a particular flow within the range There is only a FID for the whole range of port numbers and/or IP addresses.
>

GT> What prevents you from  creating another FID that is specific for that flow?

> Surprized that disclaimer for n-cast about carefull use is in the text 
> and about disclaimer for "discard" you just say "I do not think there is anything to say here." Still the question is will there be related text?
>

GT> I think the implication of dropping packets is much clearer than
the implication of n-casting. Dropping packets means the application to which the packet is sent will not receive the packet...do we really need to say that? On the other hand, receiving N copies of the same packet has some implications to TCP....this is also rather obvious for people that know TCP...but maybe not everyone does.

> Regards
> Yuri
>
>
> -----Original Message-----
> From: George Tsirtsis [mailto:tsirtsis at googlemail.com]
> Sent: Friday, October 23, 2009 11:08 AM
> To: Yuri Ismailov
> Cc: Laganier, Julien; mext at ietf.org
> Subject: Re: [MEXT] [WGLC] draft-ietf-mext-flow-binding-03
>
> Hi Yuri,
>
> Comments inline...
>
> On Tue, Oct 20, 2009 at 8:55 AM, Yuri Ismailov <yuri.ismailov at ericsson.com> wrote:
>> Hi,
>> after re-reading the draft come up with the comment on "Discard"
>> action in "Flow Identification Mobility Option"
>> I suggest either to extend the paragraph in the section 4.2 or add a 
>> separate subsection clarifying behavior of MN and HA before and after 
>> the "Discard" action was issued by MN.
>>
>> Below are the issues which are proposed to be addressed in the text.
>>
>> 1.The case when installed FID represent ranges of port numbers and IP 
>> addresses. If MN decides to discard a particular flow within the 
>> range then before that a correspondent FID should be installed in the HA.
>> Otherwise
>> "136 FID not found" will be returned from HA.
>>
>
> GT> If the MN wants to discard a particular flow within a range of
> previously registered flows, it needs to either split the range it two parts, or even better, use a higher FID-PRI for the  discard flow so that packets get matched against it before they get matched against the range.
>
>> 2. Specify the time interval for how long time the "Discard" state 
>> will be active in HA. This is necessary because there is no 
>> "un-discard" message sent from MN.
>>
>
> GT> Flow bindings share the same lifetime with the BU. There is no
> flow binding-specific lifetime. If an MN wants to remove a flow binding it sends a BU which does not include the corresponding FID.
>
>> 3. Add a disclaimer that for some protocols, like TCP, "Discard"
>> action may lead to hanging up the connection at CN and should be used responsibly.
>>
>
> GT> Discarding packets will drop connections....yes...I do not think
> there is anything to say here.
>
>>
>> Regards
>> Yuri
>>
>>
>> _______________________________________________
>> MEXT mailing list
>> MEXT at ietf.org
>> https://www.ietf.org/mailman/listinfo/mext
>>
>>
>