[ih] How long did it take?

[Tom Lyon]

In the days before TCP congestion control got a lot smarter, running TCP/IP
over slow X.25 links was extremely painful - X.25 guaranteeing to deliver
all of those already-retransmitted packets!

On Thu, Sep 17, 2026 at 2:00 PM Brian E Carpenter via Internet-history <
internet-history at elists.isoc.org> wrote:

> Yes, IP over X.25/X.75 was a thing for a while in Europe, for example when
> you couldn't get an international leased line because of PTT or funding
> agency politics.
>
> But Olivier is correct, JANET Coloured Book protocols were a better way to
> use X.25, which was a good reason why the UK transition to TCP/IP was a
> little delayed (and partly why we had to run application layer gateways for
> email and file transfer at CERN).
>
> Regards/Ngā mihi
>     Brian
> On 18-Sep-26 06:05, Craig Partridge via Internet-history wrote:
> > On Thu, Sep 17, 2026 at 11:28 AM Dave Crocker via Internet-history <
> > internet-history at elists.isoc.org> wrote:
> >
> >>
> >>> I failed to see how X.25 could be used as an internet.
> >>
> >> IP/X.25, using X.25 as a link layer.  For CSNet, Purdue wrote a Unix
> >> driver to do that.
> >>
> >> For the original CSNet email service, everything was dial-up, but that
> >> quickly included dial-up to a commercial x.28/x.29 commercial service.
> >> This meant that MMDF's packetized phonenet link layer ran on top of the
> >> commercial X.25 packet service.
> >>
> >> That is, one packet layer over another.  Purdue's driver similarly was
> >> packet over packet.  We each, separately discovered how important it was
> >> to make sure the top layer packet was no larger than the bottom layer
> >> one, if possible.  Otherwise, it doubled the commercial service's packet
> >> count, and charging was by the packet...
> >>
> >> d/
> >>
> >
> > CSNET used that Purdue code to connect a handful of CSNET sites to the
> > Internet.
> >
> > As I recall, and my memory on this is rather hazy, some of those sites
> were
> > in Europe so we went from an X.25 provider (AT&T) in the US to an X.25
> > provider in Europe.  This meant we went through an X.75 gateway that only
> > allowed, as I recall, *two* X.25 packets in flight in each direction (and
> > the pages were small - 128 or 256 bytes).  (It is useful to recall that
> > X.75 terminated the X.25 connections, so it keeping the window small was
> > important to limit buffering).  So fragmented IP datagrams over X.25
> with a
> > serious delay-bandwidth constraint.  Again, if memory serves, delays were
> > long (200ms or so).  It worked out to about 8Kbps.
> >
> > Over time, the reaction was what you'd expect: (a) customers pushed to
> use
> > premium dialup modems (the Trailblazer in the late 1980s could hit
> > 19.2Kbps); (b) dialup SLIP; and (c) leased lines.
> >
> > Craig
> >
> >
> --
> 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
>