My planned work on networking stack

[Andre Oppermann]

Brad Knowles wrote:
> 
> At 11:26 AM +0300 2004/03/02, Gleb Smirnoff wrote:
> 
> >    Is there any plans about integration of BGP routing daemon (Zebra or
> >  Quagga) into FreeBSD? With BGP routing daemon onboard, FreeBSD will be
> >  a strong alternative against expensive commercial routers. I have
> >  successfull experience of running FreeBSD STABLE with 2 full BGP views
> >  for half a year. Modern i386 PC can route/filter/shape much more traffic
> >  than expensive Cisco 36xx. I haven't yet compared with 7000 series...
> 
>         Talk to people who have real-world experience in running
> zebra/quagga in ISP environments with multiple upstreams and taking
> full views.  The guy who is designing bgpd for OpenBSD gave a talk on
> the subject at FOSDEM, and it was very enlightening to hear about the
> problems with zebra (which went commercial and the open source
> version basically hasn't been touched in years) and quagga (which is
> a community of zebra users trying desperately to fix the worst of the
> bugs), and how he has used this information during his design of a
> replacement, and the methodology he used to make sure that the
> resulting system is robust and capable of being used in real-world
> production environments.

Zebra or Quagga are not broken, just not very optimal in their
implementation.  I'm running Zebra with several full-feeds and
about 150 peerings for four years now on FreeBSD routers with
uptimes of 300-400 days.  It is true that Zebra's bgpd is un-
responsive for a couple of seconds when is has to walk the routing
table when large feeds flap but it doesn't crash.

Zebra is definatly *not* a piece of s*** as you make it sound here.

>         His only issue with using exclusively PC equipment for handling
> routing is all those strange WAN protocols and cards for which
> hardware cards are rarely available beyond vendors like cisco or
> Juniper.  That's why he's going pure Ethernet protocols/hardware
> throughout all his networks, including his upstream feeds, so that he
> can dump all that expensive ancient legacy routing hardware.

You need GigE, T1/E1, E3/T3 and STM-1 these days.  Everything else is dead.

>         If anything, I'd be inclined to look towards his work for OpenBSD
> and see if that could be imported into FreeBSD (and maybe improved,
> with contributions given back to him), rather than mess around with
> crap like zebra or quagga.

Ok, again Zebra/Quagga is not "crap".  The same with DJBware which is
no "crap" either.  If you don't like it just say so but refrain from
dirt-talking it.  It doesn't make your point any stronger.

The bgpd from OpenBSD will surely make it's way into FreeBSD [*].  The
main developer besides Henning sits about 5 meters away from me in
my office.  If you look at it then you'll find out that I'm not really
innocent that bgpd ;-)


http://www.openbsd.org/cgi-bin/cvsweb/src/usr.sbin/bgpd/rde.h?rev=1.33&content-type=text/x-cvsweb-markup

[*] In FreeBSD it will be a port.  I don't know why a bgpd should be
in the base system.

>         Oh, and it would be nice if someone somewhere started thinking
> about a mesh routing implementation for *BSD, either AODV or
> something else.

It would be nice if you could calm down, stop your mis-informed
accusations and rants and actually try to be helpful and progressive
to the projects which try to do it better.  Thank you very much.

-- 
Andre