Re: [DNA] Revised FRD (Fast Router Discovery) I-D

[Brett Pentland <brett.pentland@eng.monash.edu.au>]

Hi Syam,

Syam Madanapalli wrote:
> Hi Brett,
> 
> I am replying little late :)
> 
> On 7/19/05, Brett Pentland <brett.pentland@eng.monash.edu.au> wrote:
> 
>>Hi JinHyeock,
>>
>>I just read over your update and have a few observations.
>>
>>I see you now have triggering as well as caching.  Triggering seems
>>to make good sense when the AR and Point of Attachment (PoA) are
>>co-located, but when they are not it seems that you are just
>>reinventing RS/RA.  Compared to the host initiating the RS, you
>>save one store-and-forward time in the PoA, but that doesn't seem
>>like much of a saving for a new protocol.
> 
> 
> We have implemented the sending RS from AP while an MN is associating
> with it and found that the RA comes at MN faster than the cached RA
> sent from AP.  We have surprised with these results. It took ~700 micro 
> seconds  (0.7 ms) for an RA to reach MN from the completion of association.
> And it is very easy to implement.

I don't quite understand this.  How can sending an RS to the AR, getting
back an RA and then forwarding that RA to the MN be faster than just
forwarding a cached RA?  How much faster is it than the MN soliciting
itself?

> 
> We were sending RA from the Router kernel to MN MAC. We are sending RA
> from the router without random delay as AP sends unicast RS with TSLLAO 
> option.
> 
> A host can also send an RS but it may need to employ a some logic to avoid
> random delay.

Surely the same logic is needed for the AP or there would be collisions
if there are multiple routers on the link.

Cheers,
Brett.