Re: [dtn-interest] Re: [IRSG] POLL: draft-irtf-dtnrg-arch-07 -- DUE 30 Oct

[Scott Burleigh <[e-mail]>]

Michael Demmer wrote:
>
> On Nov 10, 2006, at 8:05 AM, Scott Burleigh wrote:
>> Having selected, potentially hazardous BRLs time out and expire 
>> sounds fine to me.  We absolutely do want to minimize the potential 
>> for production of destructive traffic.
>>
>> I'm not quite so comfortable with the concept of pre-placing BRLs 
>> along a path, in part because discovery of the path might be part of 
>> the reason to establish the BRLs.  I can certainly see how one 
>> wouldn't want BRLs for traffic-tracing BSRs installed at every node 
>> in the network, and I wouldn't argue for that.  My notion is that the 
>> subset of nodes at which we'd install some set of 
>> mutually-referencing BRLs would be selected by whoever "owned" and/or 
>> was responsible for network management at those nodes and was 
>> interested in getting specific kinds of traffic handling information 
>> from them.  Membership in the subset would be more a function of 
>> administrative interest than of topology.
>
> It seems to me that the BRL is one of potentially several ways to 
> limit the network impact of BSRs. And even with BRLs in place along 
> the entire path, a node still cannot infer definitively that failure 
> to receive a BSR means that the conditions requisite to generate the 
> BSR did not occur, as the BSR may simply have been lost by the network 
> en route to the report-to EID.
>
> I therefore propose that we change the architecture document to say 
> that nodes "should" (not "must") generate BSRs in response to the 
> specified conditions, and add a sentence or two explaining that 
> various mechanisms to rate limit or suppress unwanted BSR generation 
> may be in place at the node.
>
> Then the BRL mechanism can be individually specified in a separate 
> draft as one of these mechanisms. In addition to specifying the format 
> of the BRL and rules for the operation of the protocol, it could 
> specify that a node which conforms to the BRL spec MUST generate a BSR 
> when it has an appropriate BRL, thereby providing the additional 
> assurance that the BSR was actually created.
This seems fine to me.

Scott