Re: [Roll] [roll] #85: which lifetime is for the end points (origin and target) vs. intermediate nodes.

[Mukul Goyal <[e-mail]>]

Hi Pascal

Now we are on the same page. I propose that the following text in section 6.1 of P2P-RPL spec:

"The Default Lifetime and Lifetime Unit parameters in DODAG
      Configuration option indicate the life time of the state the
      routers maintain for a hop-by-hop route established using P2P-RPL
      and may be set as desired.

"

be replaced by the following text:

"The origin sets the Default Lifetime and Lifetime Unit parameters in DODAG Configuration option to indicate the life time of the state the routers, including the origin and the target(s), maintain for a hop-by-hop or a source route discovered using P2P-RPL.
"

Thanks
Mukul


----- Original Message -----
From: "Pascal Thubert (pthubert)" <[e-mail]>
To: "Mukul Goyal" <[e-mail]>
Cc: [e-mail], "roll" <[e-mail]>
Sent: Wednesday, April 11, 2012 1:22:52 AM
Subject: RE: [roll] #85: which lifetime is for the end points (origin and target) vs. intermediate nodes.

Mukul:

We have a disconnect because you insist in thinking that I refer to the lifetime of the temporary DAG as indicated in the RDO.
But no. I reworded twice and used very specific words this last time. 
Please remove the temp Dag from the discussion and read again in the light that I'm really talking about the states HbH routing states as I very specifically indicate.
If that does not work then we'll need a phone call to resync.

Cheers,

Pascal


-----Original Message-----
From: Mukul Goyal [mailto:[e-mail]] 
Sent: mercredi 11 avril 2012 01:59
To: Pascal Thubert (pthubert)
Cc: [e-mail]; roll
Subject: Re: [roll] #85: which lifetime is for the end points (origin and target) vs. intermediate nodes.

Hi Pascal

>I do not wish to change the DAG lifetime.

>I'm pointing out that the conversation between the origin and target cannot work longer than the lifetime of the "hop-by-hop routing states" which it depends on. So why not use that?

You mean state about the "temporary DAG" and not the hop-by-hop state for the "discovered route"? They are two different things.

>I suggest that we converge the time duration specified in DODAG configuration option to mean the maximum duration of states in all nodes that need to keep states, including origin and target, and in some case intermediate routers.

We are using the time duration in DODAG configuration option to specify the lifetime of the "hop-by-hop state for the discovered route" (and not the lifetime of the "temporary DAG"). A discovered HbH route may need to be maintained for a wide range of time durations. Hence, we would like to specify the lifetime of a discovered route using the rich semantics provided by the time duration fields inside the DODAG configuration option. On the other hand, the lifetime of a temporary DAG does not need to vary a whole lot. So, we provide 4 options for this lifetime (1, 4, 16 and 64 seconds) and specify this lifetime in a 2-bit field inside P2P-RDO.

Do you agree with how the two lifetimes (lifetime of DAG and lifetime of the discovered route) are specified in P2P-RPL?

>IOW, if the HbH route is used, the lifetime in the config option is for all states in all nodes on the path(s) including origins and target(s). If HbH routing is not used, the lifetime in the config option is still valid for the source and the origin. You may define a default value that suit classical P2P routes in the case where no config of any sort is provided. 

You seem to want the config option to specify the lifetime of the temporary DAG. This would be a wastage of rich semantic provided by those fields. If you could agree to specifying the temporary DAG lifetime inside P2P-RDO (the way P2P-RPL currently does), I see no problem whatsoever. All nodes (including the origin and target) joining the temporary DAG maintain state for the temporary DAG (which is not same as the state for the discovered HbH route) for the duration specified in P2P-RDO. We could further require that all DRO/DRO-ACK exchange MUST complete within the temporary DAG lifetime (specified inside P2P-RDO). So, the temporary DAG's lifetime has to be chosen carefully. Having said that, I still prefer giving origin and target extra time (equal to one more lifetime) to complete DRO/DRO-ACK exchange.

Thanks
Mukul

>And yes, great point, you'd need to specify that the lifetime in the config option must be large enough to accommodate the retransmissions.

>Do we converge?

>Pascal


-----Original Message-----
From: Mukul Goyal [mailto:[e-mail]] 
Sent: mardi 10 avril 2012 14:55
To: Pascal Thubert (pthubert)
Cc: [e-mail]; roll
Subject: Re: [roll] #85: which lifetime is for the end points (origin and target) vs. intermediate nodes.

OR the simpler option would be to require DRO/DRO-ACK exchange to complete within the DAG's life time? If we were to specify a new parameter for the time limit on DRO/DRO-ACK exchange, both target and origin would need to agree on the value of this parameter.

Mukul

----- Original Message -----
From: "Mukul Goyal" <[e-mail]>
To: "Pascal Thubert (pthubert)" <[e-mail]>
Cc: [e-mail], "roll" <[e-mail]>
Sent: Tuesday, April 10, 2012 7:50:16 AM
Subject: Re: [roll] #85: which lifetime is for the end points (origin and target) vs. intermediate nodes.

Hi Pascal

Hopefully we are talking about the same thing.

>No, it's not closed. We are talking about a contract between lower layers in all nodes including the source and origin to maintain necessary resources and all contracts must have a lifetime.

1. A node that joins a temporary P2P-RPL DAG maintains state for the DAG for the time duration specified as DAG life time.
2. A node that establishes hop-by-hop routing state for a route discovered using P2P-RPL maintains this state for the time duration specified in DODAG configuration option.

Origin and target need to exchange DROs and DRO-ACKs. I could specify a new configurable parameter to specify the time limit on this exchange. This parameter's value  has to be more than DAG life time. One option is to specify it in terms of existing parameters: DAG life time + (MAX_DRO_RETRANSMISSIONS + 1)*DRO_ACK_WAIT_TIME.

Thanks
Mukul
----- Original Message -----
From: "Pascal Thubert (pthubert)" <[e-mail]>
To: "Mukul Goyal" <[e-mail]>
Cc: [e-mail]
Sent: Tuesday, April 10, 2012 4:45:46 AM
Subject: RE: [roll] #85: which lifetime is for the end points (origin and target) vs. intermediate nodes.

Hi Mukul,

No, it's not closed. We are talking about a contract between lower layers in all nodes including the source and origin to maintain necessary resources and all contracts must have a lifetime.
I do not mind you overload the RPL lifetime of the routing for the states at origin and target and if you have a sentence that makes it clear then we'll be in agreement.

Cheers,

Pascal


-----Original Message-----
From: Mukul Goyal [mailto:[e-mail]] 
Sent: dimanche 8 avril 2012 17:52
To: Pascal Thubert (pthubert)
Cc: [e-mail]
Subject: Re: [roll] #85: which lifetime is for the end points (origin and target) vs. intermediate nodes.

Pascal

Can we consider this issue closed? Please see my last response.

Thanks
Mukul

----- Original Message -----
From: "roll issue tracker" <[e-mail]>
To: [e-mail], [e-mail]
Cc: [e-mail]
Sent: Wednesday, April 4, 2012 8:06:30 AM
Subject: [roll] #85: which lifetime is for the end points (origin and target) vs. intermediate nodes.

#85: which lifetime is for the end points (origin and target) vs. intermediate nodes.

 Problem (currently open)
 ------------------------------
 P2P creates temporary states in the transient DAG and less-but-still  temporary states in the endpoints. These are 2 different lifetimes. The  spec has 2 lifetimes, one in the config option and one in the RDO. The  spec is not sufficiently clear about which is which. In can appear to be  conflicting since the config option is supposed to apply to all routers  on the path. On the side, and in order to allow the reuse of instance  ID, the origin must be sure that all states for a previous usage of the  same value are gone. So we need a clear control / negotiation of the  lifetimes on the states that come with an instanceID. Again this is not  clear enough in the spec.



 [Pascal2]
 2) Same question if you want to create states at the target to route  back. How long does the target need to maintain the route? Who controls  that, the origin or the target? I'd expect to find a suggested lifetime  in the RDO with a confirmation in the DRO to let the target amend it.

 [Mukul2]
 If the target wants DRO-ACK it needs to maintain state until DRO-ACK is  received. Otherwise, the target needs to remember things until it is  done sending all the DROs. I will add the text to this effect.

 [Pascal3]
 If you are setting up a short term conversation, how long exactly is  that before the origin has to refresh the route? What controls clean up  in both sides? Usually you want a lifetime (see MIP6 for instance)

 [Mukul3]
 Is it that you are talking about the lifetime of the hop-by-hop routing  state? That is specified in the life time parameters in the DODAG  configuration option. The draft says that on page 9 while describing how  the DODAG configuration option should be set inside a P2P mode DIO:

 "The Default Lifetime and Lifetime Unit parameters in DODAG
      Configuration option indicate the life time of the state the
      routers maintain for a hop-by-hop route established using P2P-RPL
      and may be set as desired.
 "

 [Pascal4] Maybe that's so but then you need to reword a little bit cause  you got me qiute confused. I've been talking of the lifetime of the  states at origin and target for one conversation. Since they might be  having a transient conversation, and the origin might reuse the instance  ID soon, you need to give a limit in time to the states that you are  creating. Still that conversation is longer than the states in the  intermediate routers. So you have 2 lifetimes and you have to be very  clear which is which. Personally, I'd have used the lifetime in the  configuration option for all the routers on the way and I'd have used  the new lifetime in the RDO as the conversation lifetime in the end  points because:
 1)  that is the new concept.
 2) This would allow the target to confirm exactly how long it locks  resources,
 3) and this would be more compatible with the description of the config  option in RFC 6550.

 [Mukul4]
 There are two lifetimes:
 1) Lifetime of the temporary DAG: this is specified inside P2P-RDO
 2) Lifetime of the routing state for the hop-by-hop route established  using P2P-RPL: this is specified inside the DODAG configuration option.
 All routers on the established route (including the origin) maintain  state for this route for this much time. This time could very well be  infinity.

 Now, lets talk about the "states at origin and target". The origin and  the target maintain the state about the temporary DAG for the DAG's life  time. This is true for all the nodes that join this DAG. All such nodes  maintain state about the temporary DAG until the DAG's life time is  over. It is true that the target and the origin exchange DROs and  DRO-ACKs and this exchange could conceivably go on even after the  temporary DAG's demise. How long should the origin and target indulge in  this exchange (and hence remember the possibly dead DAG)? I think it is  purely their choice and the draft need not impose any artificial time  limit on this.

 Pascal

-- 
-----------------------------------+---------------------
 Reporter:  jpv@…                  |      Owner:  mukul@…
     Type:  defect                 |     Status:  new
 Priority:  major                  |  Milestone:
Component:  p2p-rpl                |    Version:
 Severity:  Submitted WG Document  |   Keywords:
-----------------------------------+---------------------

Ticket URL: <http://trac.tools.ietf.org/wg/roll/trac/ticket/85>
roll <http://tools.ietf.org/wg/roll/>