BGP Homing Question
Joe Provo
nanog-post at rsuc.gweep.net
Mon Aug 30 23:24:11 UTC 2004
[copius snips]
On Fri, Aug 27, 2004 at 11:16:40AM -0400, Patrick W Gilmore wrote:
> On Aug 27, 2004, at 8:58 AM, Joe Abley wrote:
> >On 27 Aug 2004, at 08:13, Rick Lowery wrote:
> >>I know?they would not be?good Internet citizen, but?if they needed to
> >>do this for a temp basis does anyone see an issue?
Registering everything appropriately in the IRR will help prevent
things from smelling fishy.
> It is your netblock, you get to use it as needed. This
> is much better than getting another /20 for an EU site that only needs
> a /24.
Well, for something short term it is even less complex to get
provider-allocated space. That is, you can plan the non-temporary
long-term around your PI space and have a clean transition out
of PA space. Depending on your needs -and the provider's policies-
that might be the least-disruptive path for your traffic.
> Also, filtering will not be an issue, if you are careful. Anyone who
> does not hear the /24 will hear the /20. Packets for the /24 will go
> to your US upstream.
Good advice in general for anyone concerned with more-specifics.
Reachability (and more forgiving damening) over long dstances is most
assured by making sure you are sourcing your least-specific. Lots of
networks trade more-specifics for better geographical dispersion, but
don't expect them to propagate further than those who agree to do so.
> As long as your US upstream peers with your EU
> upstream, and does not filter the /24 being announced over that peering
> link, they will send the bits where they belong. Since this is much
> more common than the alternative, you will likely have full
> connectivity.
>
> Anyone knows who filters these days?
Lots of folks; manually though? Few. Be sure your data is accurate in
[a trusted limb of] the IRR and it should be a non-issue.
> Sprint stopped when Sean left. Verio stopped when Randy left.
Tying these policies to individuals is incorrect. Sprint, NTT/Verio and
others have slid their filter windows over time, roughly in step with
RIR allocation boundaries. For example, as recently as April of this
year Verio was using /22 in classical A and B space.
The baseline expectation that the DFZ carries rechability data and any
more-specific data of interest is exchanged between parties who want it,
request it, or pay for it still holds true. "Being conservative in
what you send" also applies to anticipating *others* not being "liberal
in what they receive".
Joe
--
RSUC / GweepNet / Spunk / FnB / Usenix / SAGE
More information about the NANOG
mailing list