Re: [sig-policy] Requests from routing/packeting concerns
>
> On 17/02/2009, at 10:30 PM, Randy Bush wrote:
>
>>> Given that APNIC (and all RIRs) maintain that they cannot guarantee
>>> route-ability of any prefix assigned then I guess a case of buyer
>>> beware is appropriate.
>>
>> i just don't get it.
>>
>> why conject when you can actually test? a lot of fud and black
>> helicopter worries when one can empirically determine how much of
>> the net will be able to connect to the prefix?
>>
>> rancy
>
> That is what I was leading to. I see no value in prematurely suggesting
> a differentiation in values of like prefixes in a market when actual
> tests and searches can be done that negates the qualitative assessment.
>
I've wondered about this too and asked on our ML.
A quote from an operator in Japan:
We are aware of the testing methods, e.g ICMP test that Randy's
introduced (I've tried testing TCP connections) but there are still
some elements you can't just see from the machine test.
If you want to see reachability upto application layer for example, the
result are infinite in variety, so we can't only just on machine tests.
Let's suppose that you want to obtain address space to be used for mail
servers. I'd prefer to obtaine space that has in used by an ISP in
Japan that adopts OP25B in addition to doing a check with black list
database.
So the general sentiment by Japansese operators seems to be that;
yes, you do the machine tests to make sure that space is reachable, but
you also want to be able to see if there are other effects/risks which
you can't just see from the machine tests.
izumi