Rare project developers

[mc36 <>]

well, and to stay ontopic... i personally lost most of my interest against 
switchdev about a year ago...

here is what i have at the moment:

http://sources.freertr.org/misc/native/p4mnl_msg.h

hopefully @frederic introduced the server p4lang api (route4_add) (or you 
attended the ripe.net meeting, @fl, a recording url here thx!)



reason#1 - very poor linux support against mpls, community-wide, and as i see 
it's mostly as lack of understanding...

and time just makes it worse.... and it clearly hurts my feelings... to be 
clear, we (rare team) only uses the kernel

as a pcap device so once a random bsd offers anything 
-faster-/-bugfree-/-whatever metric-, we're happily migrating away! (alex, a 
nix link here?)

so what we use is the pcap api without the ethertype takeaways present 
nowehere else! (see rawint/pcapint/tapint in the same directory)



reason#2 - very poor emulation environment... linux and qemu only offers ipv4 
and both seems abandonware...

so clearly nor ipv6 nor mpls again....  qemu calls it rocker device 4 your 
reference, linux calls it switchdev that can program that...

(the HW offload emulator tester whatever, ip route with the magic H bit on 
the routes freerouter calculates 4 that....:))



reason#3 - very poor asic support, i mean last time i checked the kernel, 
switchdev only had open source support for the legacy mellanox asics???

so all in all, imho we're better off with something else... but change my 
mind....


reason#4 - lack ot time/trust from frederic to allow me to play with the only 
box he got some level of access...

thanks,

cs



On 4/29/23 04:07, mc36 wrote:
> so
>
> "
> wget freertr.org/rtr.zip
> unzip zzz.zip
> cd src
> ./c.sh
> ./tw.sh rout-bgp521.tst pcap r4 eth1
> " --- now gave me the attached pcap... good to see that now it dissects... 
> packet #17, #18...
>
> much easier to read except the magics (?) like why it became unknown?.... :)
>
> seeked also into the rfc, imho im right here?
>
> (tbh, i have the feeling, that 2021-12 was too early for my skills, and maybe 
> others are earlier implementers?!... :)
>
> my google also failed me to find the mentioned pdf about the 'slides', an url 
> would be nice from you, please....
>
>
>
> but, im still scratching my head on things like why the cicd went crazy about 
> the yesterday commit??? any help highly appreciated that.... :)
>
> 6he update will pass tomorrow imho against the big vendors as i already 
> started using it internally (inf.nop.hu) and niif.hu updates 10am
>
> today (telnet lg.hbone.hu to see this on the full table plus the local ix's 
> blackhole /32s, plus the geant mdvpn, etc....:),
>
> then slowly the whole lab, including frederic daily morning schedule... and 
> so on.... not to mention my dn42.dev (telnet sandbox.freertr.org) peering box 
> and internal bgp (lg.nop.hu)
>
> so what i saw is that r1 refused the prefix from r4, according the config, i 
> think thats a way better behavior than the 100% success rate???
>
> regrading the frr help, so basically i just meant the core lines like 
> remote-as and the role.... im asking as the frr docs now confused
>
> me a little bit, it seems an industry standard split here, and as we test 
> against intop9-frr.htm, im also happy revert the yesterday changes...
>
> but i have the feeling i should not, just wait for the cicd to fail and then 
> fix the test case then.... :)
>
> thanks,
> cs
>
>
>
>
> On 4/29/23 00:14, mc36 wrote:
> > On 4/28/23 23:46, mc36 wrote:
> > > also git 3 dots to spot the auto-shitload when you do edit the project's 
> > > default todo.txt instead of the issue tracker...
> >
> > https://github.com/rare-freertr/freeRtr/blob/master/changelog.txt#L12322