[ih] X.25

Jack Haverty jack at 3kitty.org
Mon Sep 29 12:48:35 PDT 2025


Lots of info here:

https://apps.dtic.mil/sti/tr/pdf/ADA137427.pdf   (See especially section 
1.1.2)

and here:

https://apps.dtic.mil/sti/tr/pdf/ADA206354.pdf

I don't recall if X.25 was ever used in the ARPANET, or in the public 
Internet as a way to interface to host computers.

Sometime prior to 1983 we had added an X.25 interface to the core 
gateway at BBN, which enabled it to use the public X.25/X.75 network as 
an alternate path across the Atlantic to provide connectivity between 
the US and Europe.  There was also a gateway at UCL with an X.25 
interface.  As far as I remember there were never any host computers 
using X.25 to access those gateways, but the two gateways could interact 
over the X.25 world just as they did over other networks.

The motivation for that so-called "VAN Gateway" was to allow projects in 
EU that were not ARPA-related to have connectivity to US sites on the 
ARPANET.  ARPA didn't want the SATNET resources (primarily the Intelsat 
IV-A satellite channel) to be used for non-ARPA projects.

More info here: https://apps.dtic.mil/sti/tr/pdf/ADA122596.pdf (see page 35)

At the time, the routing mechanisms of the Internet were unable to 
handle such "policy routing", but could only choose a route based on hop 
count.  IIRC, we had to use a few IP tricks to make the different types 
of traffic take the desired routes.  I think that was accomplished by 
giving some LANs in Europe two different network numbers.   To connect 
to a host on that LAN using the SATNET route, you would use an IP 
address on the "non-ARPA" network.  To connect to a host using the X.25 
route, you would use an IP address on the ARPA network.

It was a kludge, but it worked.  It also put "policy routing" on the 
ICCB's to-do list for the next generation of routing protocols, and 
validated that X.25 could be used within the TCP/IP universe.

Jack Haverty




On 9/29/25 12:07, Michael Grant via Internet-history wrote:
> X.25 has been mentioned a few times on this list in the context some 
> IMPs could talk to other IMPs over X.25 and somehow parts of the net 
> (arpanet?) were connected over X.25.
>
> X.25 addresses were sort of like phone numbers, they don't map on to 
> IP addresses (or IMP/HOST).  And data was charged per packet, per 
> byte, and iirc, per connection time.  And it was connection oriented 
> though there was a sort of datagram mode called "fast connect".
>
> How was it used in the early Internet?  Was there some static file 
> passed around with how to route to something and which X.25 address to 
> connect to?  Or was each "link" to other IMPs treated as a point to 
> point connection with some configured X.25 address?
>
> When I was working with the OSI protocols, X.25 was somehow supposed 
> to be used as a network layer but I never could imagine how that was 
> supposed to work in any practical way.  CLNP has a really long address 
> that you could embed an X.25 address in but it seemed senseless (if 
> you used CLNP). Who would be paying for that if something got routed 
> over your link!  It would cost you a fortune!
>
> So I'm curious, how did X.25 fit into things in the early internet?  
> Was it used much?  I just never could understand how X.25 was anything 
> like the Internet in the OSI world if you had an X.25 connection.
>
> Michael Grant

-------------- next part --------------
A non-text attachment was scrubbed...
Name: OpenPGP_signature.asc
Type: application/pgp-signature
Size: 665 bytes
Desc: OpenPGP digital signature
URL: <http://elists.isoc.org/pipermail/internet-history/attachments/20250929/502481e0/attachment-0001.asc>


More information about the Internet-history mailing list