Firstly, I must note that there are a whole ot of use of "may" as distinct from 
the normative "MAY" in this document, "should" as distinct from the normative 
"SHOULD", etc.  As thisd is a document intended for a Standards Track, the 
document needs to be strict about its use of the normative forms of the 
requirements language as noted in section 1.1 of the document


e.g.: " An AS _may_ originate more than one prefix set.  Thus, multiple prefix 
sets in the database _may_ contain the same origin AS(es).


On 01/12/2010, at 4:59 AM, Pradosh Mohapatra wrote:

> Hi Geoff,
> 
>> Section 5. Routing Policy
>> 
>>  Announcements with invalid origins MAY be used, but SHOULD be less
>>  preferred than those with valid or unknown.
>> 
>> 
>> Here I agree with the sentiment in terms of the relativity to valid or 
>> unknown, but I am worried about the MAY statement here, and I would like the 
>> document to explain this further.
>> 
>> For example, a simple reading of the document leads one to believe that all 
>> announcements MAY still be accepted according to the routing policy 
>> selection allowed in section 5, yet if that were the case then the ability 
>> of origin validation to "deal with inadvertent mis-advertisement" is 
>> questionable.
>> 
>> I would prefer the document to look at the implication of _why_ a party MAY 
>> accept a route that the origination validation framework is leading to a 
>> local judgement that the route is invalid. i.e. is it because of missing 
>> information in the relying party's local cache, which may bne resolved over 
>> time? Is is due to potential circularity of dependence, where parts of the 
>> RPKI distributed repository lie behind routes that can only be judged as 
>> valid using certs and ROAs found in that repository? 
> 
> I am afraid that any attempt to enumerate the cases where an invalid route 
> MAY be accepted will be incomplete! Rather, could we add a statement to the 
> effect that accepting invalid routes is a pure local matter and should be 
> done with utmost care?


works for me

  Geoff

_______________________________________________
sidr mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/sidr

Reply via email to