[ih] Overlay networks

Jack Haverty jack at 3kitty.org
Sat Aug 23 11:34:19 PDT 2025


At some point in the late 70s, I remember testing out some TCP 
implementation, trying to evaluate how many connections could be 
simultaneously supported.  To do that, I logged into machine#1, opened a 
telnet connection to machine #2, logged in, opened a telnet connection 
back to machine #1, and repeated such looping to see when it would no 
longer allow another connection.  I think the machines were running Tenex.

To interact with and control Telnet, you had to type an escape character 
(can't remember what it was).  To actually transmit an escape character 
to the other end of the connection, you had to type it twice.  To 
interact with the Nth telnet connection in the loop, you therefore had 
to type 2^N escape characters.

At some point I lost count of how many escape characters I had typed.   
I could still tell which machine I was talking to, but not which 
connection I was controlling.

IIRC, getting that mess cleaned up required the system operators to 
intervene.

At some point in the 1970s, I remember accidentally creating a "loop" in 
the email network.  It was amazingly simple to do.  All that was 
required was to add an address to some mailing-list, where the new 
address was actually the address of another mailing list on some other 
machine.   With the emerging fascination with mailing lists at the time, 
it was easy to imagine how a complex loop of emails might be created.

Also in the early 1990s, while I was at Oracle, we purposely created an 
overlay network, which enabled clients and database servers to interact, 
even id they were on different kinds of networks.  For example, a client 
PC might be on a LAN using Netware, using a database on an IBM mainframe 
using SNA (LU6.2).  The technology was analogous to the Internet's 
approach to connecting different types of networks using gateways (now 
routers).   Our system was called TNS (Transparent Networking Substrate) 
and the interconnection points (software only) were called 
Interchanges.   It provided an alternative to architectures using 
multiprotocol routers.  Our customers found the alternative much more 
palatable to operate.

I'm not sure if any of those scenarios would be called "overlays". But 
they did happen, and I wonder if scenarios such as the email one are 
still possible today, such as in forums like this one.

Jack



On 8/23/25 11:04, Karl Auerbach via Internet-history wrote:
> Thinking of overlay networks...
>
> Back in the late 1980s and early 1990s there were a bunch of small 
> companies who played and partied (and I mean *seriously* played and 
> partied) together.  These included FTP Software, TGV, Intercon, 
> Internode, Beame & Whitside, Epilogue Technology, Empirical Tools and 
> Toys.  (Epilogue and Empirical were companies I founded.  I learned 
> later that I had been an indirect cause of the formation of FTP 
> Software.)  From this group came the core of the shownet team that 
> designed and deployed the Interop show networks - we got really good 
> at deploying convention-center + city-spanning multi-protocol networks 
> of thousands of nodes within a span of a few hours (as short as eight 
> hours.)
>
> So it fit right in when Stuart Vance (TGV) and Simon Hacket 
> (Internode) wanted to play a kind of network Jinga.
>
> So they created an overlay network - link between TGV (Santa Cruz, 
> California) and Internode (Adelaide, Australia.)
>
> This wasn't just any link.  It was a stack of various protocol 
> families including IPv4, IPX/Netware, and Decnet.  I can't remember 
> whether IP, Netware, or Decnet was the bottomost at any given moment.  
> And sometimes there were multiple instances, like IP over Decnet over 
> Netware over Decnet over IP.
>
> Surprisingly, connections at the topmost layer could be created and 
> sustained, although the round trip time was rather long.
>
> As I said, we like to play.  We did the first Internet toasters, two 
> distinct implementations.  (We licensed the Berkeley Designs image of 
> flying, winged toasters for the celebratory T-shirt.)  We did 
> "Etherphones", essentially an early VoIP (the T-shirt for that is a 
> classic, based on the Sistine Chapel, with God handing a phone to 
> Adam.)  We also had a stereo system, with a 100 slot CD-player/library 
> located in Santa Cruz that Simon could feed and control from 
> Adelaide.  There was also the Internet toaster-loading Lego robot, a 
> talking bear, and a weather station.  We also had contact with The 
> Little Garden network (I remember being at the lunch where everyone 
> met Little Garden restaurant in Palo Alto to get things going.)
>
> And if I remember correctly, we served as forwarding/delivery-dialing 
> nodes on Marshall Rose/Carl Malamud's Internet based overlay faxing 
> network, TPC.INT - RFC 1528 - https://datatracker.ietf.org/doc/rfc1528/
>
>         --karl--
>
>
>

-------------- 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/20250823/954919c9/attachment-0001.asc>


More information about the Internet-history mailing list