[ih] History question -- multi-homing
John Shoch
j at shoch.com
Thu Sep 10 17:12:43 PDT 2026
In a recent post John Day raised an interesting question on
multi-homing. There were a couple of quick historical replies and I
was also going to offer a bit of aditional history, but then realized
the original question raises a bunch of complicated issues.
A. Some thoughts:
1. In the simple case, multi-homing means a device with multiple
network connections. We usually think of this as connections to
different networks, but it could also be multiple connections within
one network (a device in a cell phone network is multi-homed to
different cell sites, switching dynamically).
Yet the broader question is, what do you try to do with it?
2. At the edge, or for a client, multi-homing usually provides some
redundancy or failover.
I think my first home internet connection was a dial-up modem.
Then I moved up to ADSL, but kept the modem and became multi-homed.
Failover required manually moving a cable!
I then moved up to a cable internet connection, keeping the ADSL line
for backup. I think we used a PIX router, to help re-configure when
needed.
Today I sit with a Mac on both an Ethernet and WiFi; if either of
those fails, the Mac takes care of moving my traffic via the other
path (brilliant).
3. At a server multi-homing can also provide improved service or performance.
Network-capable printers these days can receive files from your
desktop (via an Ethernet) or from your phone (via WiFi).
In very large deployments a server can have multiple network
interfaces to improve parallel service to multiple clients.
But things get more complicated with naming, addressing, and routing.
Some designs would let you address a specific interface in a host (for
testing, or because you will get better performance via a higher-speed
access point). But if that path fails you will need a mechanism to
find a new address.
Some designs would let you address the host, and the routing mechanism
picks the path to a particular interface. If that path fails, a
dynamic routing process should automatically pick another route to the
host.
Much more ambitious, though, would be an effort to use multi-homing to
load-balance -- splitting one traffic stream across two different
paths to a multi-homed host.
So, to discuss "multi-homing" we may need a bit more context -- what
problem are we trying to solve?.
B. Some history
1. Earlier this year I sent around a note about the state of our
internet operating at Xerox PARC in March 1976, which included three
full-fledged gateways connecting multiple networks (multiple
Ethernets, multiple Data General MCAs, and the Arpanet).
2. We could have had one host with two Ethernet interfaces connected
to the same Ethernet, although I am not sure what we would have done
with it.
3. More importantly, there were other dedicated servers (not
gateways) that were multi-homed, to provide service to machines on
different networks (either an Ethernet or a Data General MCA):
--MAXC, running Tenex (similar to what Bill Yeager described 6 years
later at Stanford).
--XGP CRT print server
--EARS laser print server
--The Woodstock page-at-a-time file server.
4. The internetwork gateways were, of course, multi-homed and
provided decapsulation, encapsulation, routing, etc. for different
networks.
Cheers,
John S.
On Mon, Sep 7, 2026 at 11:35 AM
<internet-history-request at elists.isoc.org> wrote:
>
> Send Internet-history mailing list submissions to
> internet-history at elists.isoc.org
>
> To subscribe or unsubscribe via the World Wide Web, visit
> https://elists.isoc.org/mailman/listinfo/internet-history
> or, via email, send a message with subject or body 'help' to
> internet-history-request at elists.isoc.org
>
> You can reach the person managing the list at
> internet-history-owner at elists.isoc.org
>
> When replying, please edit your Subject line so it is more specific
> than "Re: Contents of Internet-history digest..."
>
>
> Today's Topics:
>
> 1. History question (John Day)
> 2. Re: History question (Bill Woodcock)
> 3. Re: History question (Craig Partridge)
> 4. Re: History question (Dave Crocker)
>
>
> ----------------------------------------------------------------------
>
> Message: 1
> Date: Mon, 7 Sep 2026 13:38:42 -0400
> From: John Day <jeanjour at comcast.net>
> To: internet-history at elists.isoc.org
> Subject: [ih] History question
> Message-ID: <C43A7A54-5783-4F01-857E-62EDBE51247A at comcast.net>
> Content-Type: text/plain; charset=us-ascii
>
> When and who brought the multihoming question to general discussion?
>
> When was it first recognized as an issue, even if there was no real discussion about a solution?
>
> Take care,
> John Day
>
> ------------------------------
>
> Message: 2
> Date: Mon, 7 Sep 2026 20:04:27 +0200
> From: Bill Woodcock <woody at pch.net>
> To: internet-history at elists.isoc.org
> Subject: Re: [ih] History question
> Message-ID: <2ad11dd9-f274-41ca-8f1f-0e84ec3044d3 at pch.net>
> Content-Type: text/plain; charset="utf-8"; Format="flowed"
>
> On 9/7/26 12:38, John Day via Internet-history wrote:
> > When and who brought the multihoming question to general discussion?
> >
> > When was it first recognized as an issue, even if there was no real discussion about a solution?
>
> Assuming you mean the multihoming of using multiple "upward" direction
> networks to ensure continuous availability of transit, that was already
> established practice when I built my first one in 1989.
>
> So it predated the modern NII / IXP architecture and the transit /
> peering distinction. So at that time there was no clear direction "up"
> but more just generally "backboneward" since the backbone didn't have a
> specific center, and multihoming didn't necessarily mean toward that
> center because transit cones hadn't yet been defined.
>
> BGP was standardized in 1989. EGP as a subset of GGP in 1982.
>
> https://www.rfc-editor.org/rfc/rfc823.txt
> https://www.rfc.fr/rfc/en/rfc0827.pdf
> https://www.rfc-editor.org/info/rfc1105/
>
> -Bill
> -------------- next part --------------
> A non-text attachment was scrubbed...
> Name: OpenPGP_signature.asc
> Type: application/pgp-signature
> Size: 840 bytes
> Desc: OpenPGP digital signature
> URL: <http://elists.isoc.org/pipermail/internet-history/attachments/20260907/78782fea/attachment-0001.asc>
>
> ------------------------------
>
> Message: 3
> Date: Mon, 7 Sep 2026 12:26:17 -0600
> From: Craig Partridge <craig at tereschau.net>
> To: John Day <jeanjour at comcast.net>
> Cc: internet-history at elists.isoc.org
> Subject: Re: [ih] History question
> Message-ID:
> <CAHQj4CdLy=kA1FnDLv8CP7mhMfrX=UHgCsfFw+PM_aFW1w3=4Q at mail.gmail.com>
> Content-Type: text/plain; charset="UTF-8"
>
> Are you talking about multihoming a host or a network (host connected to
> multiple networks, network connected to multiple providers).
>
> For hosts, the issue appeared c. 1982/3 when Berkeley connected UCBVAX to
> multiple Ethernets (I think for reliability reasons). Around the same time
> they also put IP router capability in the BSD networking code. Berkeley
> thought deeply about the issues but, for reasons unclear, didn't surface
> their insights in the community until the late 1980s. I remember Mike
> Karels at an IETF Host Requirements meeting, going through a late draft of
> the requirements document. At various points he said "you don't want to say
> that as it breaks multihoming" and then had to explain to the room what the
> issue was, because most of us were clueless.
>
> For networks, the advent of NSFNET in 1987. Suddenly you could be
> connected to both ARPANET and NSFNET. Also, because NSFNET was flakey,
> people leased direct lines to peer institutions. EGP promptly displayed
> all its limitations -- there's a SIGCOMM '87 paper where Dave Mills and
> Hans-Werner Braun presented interim solutions.
>
> Craig
>
> On Mon, Sep 7, 2026 at 11:38?AM John Day via Internet-history <
> internet-history at elists.isoc.org> wrote:
>
> > When and who brought the multihoming question to general discussion?
> >
> > When was it first recognized as an issue, even if there was no real
> > discussion about a solution?
> >
> > Take care,
> > John Day
> > --
> > Internet-history mailing list
> > Internet-history at elists.isoc.org
> > https://elists.isoc.org/mailman/listinfo/internet-history
> > -
> > Unsubscribe:
> > https://app.smartsheet.com/b/form/9b6ef0621638436ab0a9b23cb0668b0b?The%20list%20to%20be%20unsubscribed%20from=Internet-history
> >
>
>
> --
> *****
> Craig Partridge's email account for professional society activities and
> mailing lists.
>
>
> ------------------------------
>
> Message: 4
> Date: Mon, 7 Sep 2026 18:35:01 +0000 (UTC)
> From: Dave Crocker <dhc at dcrocker.net>
> To: Craig Partridge <craig at tereschau.net>
> Cc: internet-history at elists.isoc.org
> Subject: Re: [ih] History question
> Message-ID: <44150783-615c-4d0e-9d41-95ec51b6eb43 at dcrocker.net>
> Content-Type: text/plain; charset=UTF-8; format=flowed
>
> On 9/7/2026 11:26 AM, Craig Partridge via Internet-history wrote:
> > For networks, the advent of NSFNET in 1987. Suddenly you could be
> > connected to both ARPANET and NSFNET.
>
>
> I had heard -- though I have no details -- that there was ad hoc
> multi-homing of networks prior to that, though NSFNet certainly forced
> formalization of the capability.
>
> d/
>
> --
> Dave Crocker
>
> dhc at dcrocker.net
> bluesky: @dcrocker.bsky.social
> mast: @dcrocker at mastodon.social
> +1.408.329.0791
>
> Volunteer, Silicon Valley Chapter
> Northern California Coastal Region
> Information & Planning Coordinator
> American Red Cross
> dave.crocker2 at redcross.org
>
>
>
> ------------------------------
>
> Subject: Digest Footer
>
> Internet-history mailing list
> Internet-history at elists.isoc.org
> https://elists.isoc.org/mailman/listinfo/internet-history
> -
> Unsubscribe: https://app.smartsheet.com/b/form/9b6ef0621638436ab0a9b23cb0668b0b?The%20list%20to%20be%20unsubscribed%20from=Internet-history
>
>
> ------------------------------
>
> End of Internet-history Digest, Vol 82, Issue 1
> ***********************************************
More information about the Internet-history
mailing list