On Fri, Aug 26, 2011 at 03:24:54PM +0100, Stuart Henderson wrote: > Some networks use AS23456 in their IRR macros because they have > downstreams with ASN >= 65536 and obviously they want any > 16-bit-only-speaking upstreams to accept their customer prefixes. > I noticed AS-BLANK and AS-SPACENET but there are others; see > http://www.robtex.com/as/as23456.html > > So currently, for macros which end up including 23456, bgpctl > irrfilter generates filter files which bgpd won't read. > > Would it make sense for irrfilter to avoid including these > (possible diff below) or should bgpd accept this in filters? >
Honestly hard to tell. I think there may be a valid point to be able to filter on 23456 so maybe bgpd should accept it for filters. I think the unix rule of providing as much rope as possible could make sense here. > Index: irr_output.c > =================================================================== > RCS file: /cvs/src/usr.sbin/bgpctl/irr_output.c,v > retrieving revision 1.13 > diff -u -p -r1.13 irr_output.c > --- irr_output.c 5 Mar 2007 17:28:21 -0000 1.13 > +++ irr_output.c 26 Aug 2011 14:20:15 -0000 > @@ -143,8 +143,9 @@ policy_torule(FILE *fh, struct policy_it > if (strlen(ass->as[i]) < 3) > errx(1, "%s not AS...", ass->as[i]); > srcas = ass->as[i] + 2; > - for (j = 0; j < pfxs->prefixcnt; j++) > - print_rule(fh, pi, srcas, pfxs->prefix[j]); > + if (!strcmp(srcas, "23456")) > + for (j = 0; j < pfxs->prefixcnt; j++) > + print_rule(fh, pi, srcas, > pfxs->prefix[j]); > } > } > } > -- :wq Claudio
