I'd be interested in hearing if people set a hostname on their BGP routers and send it to peers (internal and/or external). I'm also interested in hearing if you see external peers sending it to you. I know BIRD and FRR support this capability, but I"m not aware of others and I'm guessing it is relatively rare in practice? Reference: <https://datatracker.ietf.org/doc/html/draft-walton-bgp-hostname-capability-02> Yes, a long-expired draft. John
Hi. Personally, I have never seen this in practice. While reading the "Hostname Capability for BGP" draft, my immediate thought was that perhaps this could be done simply in the user interface using rDNS. On the other hand, I guess there are some cases like BGP sessions between IPv6 link local addresses when the rDNS approach would not work. Martin
On 11/12/25 14:14, John Kristoff via NANOG wrote:
I'd be interested in hearing if people set a hostname on their BGP routers and send it to peers (internal and/or external).
I'm also interested in hearing if you see external peers sending it to you.
I know BIRD and FRR support this capability, but I"m not aware of others and I'm guessing it is relatively rare in practice?
There might be some corner cases where it makes sense, but if you're trying to bring up a BGP router on the Internet and it can't reach DNS until it's online you're going to have some issues. -- Jay Hennigan - jay@west.net Network Engineering - CCIE #7880 503 897-8550 - WB6RDV
Den 12.11.2025 23:14, skreiv John Kristoff via NANOG:
I'd be interested in hearing if people set a hostname on their BGP routers and send it to peers (internal and/or external).
I'm also interested in hearing if you see external peers sending it to you.
Yes and yes. We see a few route server operators (Netnod and Verizon) to be specific advertise it to us. We find it useful, as it makes the neighbour summary more human readable for those peers, as we do not have to remember their IP addresses and/or AS numbers to know which peer it is at a glance: $ show bgp ipv4 unicast vrf internet summary … r1.nix.dnadm.se(185.1.55.200) 4 8674 2289433 2289357 0 0 0 04w2d03h 0 4 r1.osl.dnadm.se(185.1.55.201) 4 8674 2289420 2289341 0 0 0 02w1d02h 0 4 athenalb1.elosl3.electra.vrsn.com(185.1.55.204) 4 26415 2286083 2286043 0 0 0 01w0d18h 0 4 Tore
On Thu 13 Nov 2025 13:25:10 GMT, Tore Anderson via NANOG wrote:
Den 12.11.2025 23:14, skreiv John Kristoff via NANOG:
I'd be interested in hearing if people set a hostname on their BGP routers and send it to peers (internal and/or external).
I'm also interested in hearing if you see external peers sending it to you.
Yes and yes. We see a few route server operators (Netnod and Verizon) to be specific advertise it to us. We find it useful, as it makes the neighbour summary more human readable for those peers, as we do not have to remember their IP addresses and/or AS numbers to know which peer it is at a glance:
$ show bgp ipv4 unicast vrf internet summary
…
r1.nix.dnadm.se(185.1.55.200) 4 8674 2289433 2289357 0 0 0 04w2d03h 0 4 r1.osl.dnadm.se(185.1.55.201) 4 8674 2289420 2289341 0 0 0 02w1d02h 0 4 athenalb1.elosl3.electra.vrsn.com(185.1.55.204) 4 26415 2286083
I also see it on some eBGP sessions on the RS as an IXP operator. On my side, I use it on iBGP on my RRs as I configured a range to be the dynamic peers, and I know who’s who with it. -- Alarig
On Wed, 12 Nov 2025 16:14:10 -0600 John Kristoff <jtk@dataplane.org> wrote:
I'd be interested in hearing if people set a hostname on their BGP routers and send it to peers (internal and/or external).
I'm also interested in hearing if you see external peers sending it to you.
I know BIRD and FRR support this capability, but I"m not aware of others and I'm guessing it is relatively rare in practice?
Most of what I learned about usage was in response to this thread, but I had a few brief responses from others elsewhere. Awareness and interest in it remains lukewarm at best. I soon after realized that ExaBGP supported this capability as well. The original discussion and reaction (pro and con) to the feature is here: <https://mailarchive.ietf.org/arch/search/?q=%22draft-walton-bgp-hostname-capability%22> Also note, there is an IANA assignment for it (73). <https://www.iana.org/assignments/capability-codes/capability-codes.xhtml> I'm guessing this capability will just sort of live on forever, rarely used until, if ever, those software implementations decide they want to remove it. John
Hey John, Thus spake John Kristoff via NANOG (nanog@lists.nanog.org) on Mon, Nov 17, 2025 at 10:53:43AM -0600:
On Wed, 12 Nov 2025 16:14:10 -0600 John Kristoff <jtk@dataplane.org> wrote:
I'd be interested in hearing if people set a hostname on their BGP routers and send it to peers (internal and/or external).
I'm also interested in hearing if you see external peers sending it to you.
I know BIRD and FRR support this capability, but I"m not aware of others and I'm guessing it is relatively rare in practice?
Most of what I learned about usage was in response to this thread, but I had a few brief responses from others elsewhere. Awareness and interest in it remains lukewarm at best.
I soon after realized that ExaBGP supported this capability as well.
The original discussion and reaction (pro and con) to the feature is here:
<https://mailarchive.ietf.org/arch/search/?q=%22draft-walton-bgp-hostname-capability%22>
Also note, there is an IANA assignment for it (73).
<https://www.iana.org/assignments/capability-codes/capability-codes.xhtml>
I'm guessing this capability will just sort of live on forever, rarely used until, if ever, those software implementations decide they want to remove it.
I have to admit I didn't even know it existed. In our measurement/monitoring systems we do record the BGP router ID of every neighbor. This is handy as it should[*] be unique to an AS and can help with troubleshooting for where you may have multiple peering sessions to the same router, particularly across address-families. IMHO, adding a hostname field seems at best duplicative and probably is just another thing to be poorly maintained and go stale over time. Dale [*] that should is lower-case in RFC6286 section 2.1 ...huh
Thus spake Randy Bush (randy@psg.com) on Tue, Nov 18, 2025 at 02:55:07PM +0900:
In our measurement/monitoring systems we do record the BGP router ID of every neighbor. This is handy as it should[*] be unique to an AS
uh, my memory is that the spec is that it is unique WITHIN an AS. two ASs can have routers with the same routerID.
doh! yes, typo on my part. Dale
This is comparatively new change in RFC6286 (2011, which in RFC time is last weekish) Prior to RFC6286 router-id had to be globally unique. On Tue, 18 Nov 2025 at 15:22, Dale W. Carder via NANOG <nanog@lists.nanog.org> wrote:
Thus spake Randy Bush (randy@psg.com) on Tue, Nov 18, 2025 at 02:55:07PM +0900:
In our measurement/monitoring systems we do record the BGP router ID of every neighbor. This is handy as it should[*] be unique to an AS
uh, my memory is that the spec is that it is unique WITHIN an AS. two ASs can have routers with the same routerID.
doh! yes, typo on my part.
Dale _______________________________________________ NANOG mailing list https://lists.nanog.org/archives/list/nanog@lists.nanog.org/message/VWKDZ6WT...
-- ++ytti
On Tue, 18 Nov 2025 at 16:42, Dale W. Carder <dwcarder@es.net> wrote:
Prior to RFC6286 router-id had to be globally unique.
This must have only been implicit as RFC 1771 is < 1918?
I may not understand what is intended to communicate here. But explicitly RFC1771 needed a globally unique router-id to handle collision on open and some other reasons. RFC6286 modifies these behaviours. -- ++ytti
I'm guessing this capability will just sort of live on forever, rarely used until, if ever, those software implementations decide they want to remove it.
Considering that router IDs may be arbitrary numbers, I expect more hostnames to be seen on the BGP sessions. In a network where the BGP links establish more or less automagically over link-local addresses, all the legacy stuff uses RFC 8950 and router ids are more or less random, hostname is the only reasonable thing to set to identify a host well. Also there are still 15+yo implementations out there in the wilderness, you can't expect people to eagerly update everything just to use the hostname capability. Its use will simply gradually increase as new deployments roll out. Last but not least, it doesn't help at all that this is a long expired draft, and it deserves to become an RFC. Maria -- Maria Matejka (she/her) | BIRD Team Leader | CZ.NIC, z.s.p.o.
participants (9)
-
Alarig Le Lay -
Dale W. Carder -
jay@west.net -
John Kristoff -
Maria Matejka -
Martin Tonusoo -
Randy Bush -
Saku Ytti -
Tore Anderson