[as112-ops] LOA Request for AS112 Prefixes
Frank Habicht
geier at geier.ne.tz
Fri Nov 2 18:04:45 UTC 2018
Hi Joe,
On 02/11/2018 20:21, Joe Abley wrote:
> Hujambo rafiki!
>
> On Nov 2, 2018, at 12:46, Frank Habicht <geier at geier.ne.tz> wrote:
>
>> Of course: many good things have come from volunteering things. But here
>> one is also opening the possibility for making flows less efficient, if
>> say your upstream now gets AS112 from one or many peerings, and after
>> this change they get and prefer it from you (customer) and then send
>> traffic from around the world to you...
>>
>> I just believe AS112 packets shouldn't travel far.
>
> I agree with you, but for the service to be globally available
> *someone* needs to provide transit for the service addresses.
absolutely yes.
I think there are some usual suspects (HE, or am i confusing with
192.88.99/24?) and (iirc) some (well anycasted) root operators...?
those who "don't buy transit" could (should) run their instances, or
peer with AS112 without telling anyone.
Just sayin' that it's worth thinking twice about it.
If not for the traffic, then for latency from far away.
In the past, having been connected to a "network that doesn't buy
transit" in UK, that sent traffic to <letter>.root-servers.net to
Thailand, just because that one was a customer (or customer's customer)
... left me unimpressed.
> In the past I have used AS112 as a test tenant on new anycast
> infrastructure in order to do end-to-end testing. It's a service with
> live traffic and no reasonable end-user expectations, so it's kind of
> perfect for that.
>
> Whether you or I think that these routes *should* be advertised for
> transit, I think it would be unreasonable to insist that they should
> not.
Yep. I'm not insisting they don't. Am suggesting there are (many) cases
where one should think twice.
Who would want their resolver to ask an auth far away for such an
important query.... ;-)
Frank
More information about the as112-ops
mailing list