On 11.09.2009, at 10:27, Ondrej Filip wrote:

On 11.9.2009 08:58, Wolfgang Hennerbichler wrote:

...

Good Morning Ondrej,

Hi Wolfgang! great news! However please wait for next release. I fixes one more BGP related bug.

Hi Ondrej, thanks for answering. before we go into production maybe 2 months or so will pass, so no problem, I guess the next release will be out before us going productive :)

...

I've got bird running in a test-setup where it's already working great as a route-server with an extremely low memory footprint. Grats for that! I'm also getting more and more confident in working with, understanding and using bird. However a question that came to my mind is the strategy you chose here:

On 21.08.2009, at 22:50, Ondrej Filip wrote:

...

All session are unfiltered (except max prefix limit) and connected to routing table named T<AS#>. So it means all prefixes received from R25192x1 and R25192x2 are store in routing table called T25192. Each AS has its own routing table. All those routing table are connected to the master RT by pipe protocols named P<AS#>. Filtering (both in and out) is applied on those PIPE protocols.

Why don't you filter in the BGP sessions directly instead of in the pipe-protocol, this saves you some precious memory and doesn't do any harm, or am I wrong in my thinking?

That is because I wanted to know what the IXP members sent to me before filtering. I want them to have good export filters, so I compare difference before and after filtering. That should be the same in optimal scenario.

I see. So this is your way of 'soft reconfiguration'.

But you way is correct as well of course.

Good to know :)

Ondrej

...

Wolfgang

-- ( CZ.NIC z.s.p.o. ) ------------------------------------------------- Ondrej Filip - CEO

Office : Americka 23, Praha 2, Czech Republic Email : ondrej.filip@nic.cz http://www.nic.cz Private: feela@network.cz -------------------------------------------------

-- www.vix.at | www.aco.net wh@univie.ac.at | WH844-RIPE Vienna University Computer Center