[ih] Speaking of layering and gateways

[Jack Haverty]

Good summary!  That's a description of the techniques used and problems 
encountered *within* the "railway networks".   Perhaps railroad gauge 
diversity in the 19th century was recreated as multiprotocol routers in 
the 20th.   Both were abandoned as soon as possible.

But rails didn't usually get all the way to the sources and 
destinations.    In the 1800s, for example there were "team tracks", 
where a railcar could be placed by a platform with a road on the other 
side - where you would park your team of horses and their wagon for the 
next, or previous, leg of the journey over a road or cartpath network.   
Today we have container terminals, where cranes lift containers between 
rail cars and trucks.  We might call both of these "gateways"...

I also remember thinking about the economic aspects of rail transport.  
Cargo was often charged by weight and distance.  But some railroads 
charged different amounts for different cargoes even of the same 
weight.  That practice is even captured in a folk song: 
https://www.johnnycash.com/track/rock-island-line-3/

Of course there were many distinct railroad companies, each with their 
own network, and sometimes cargo had to traverse multiple such 
networks.   Railroads met at "junctions", often triggering the creation 
of a surrounding town - e.g., Colorado Junction.    There were rules for 
operation of such junctions, mutually agreed by the railroads 
involved.   In today's Internet, we call such places IXPs.

There were complex agreements, rules, and regulations developed over 
time.  For example, railroad cars full of goods generated revenue for 
the rails carrying them.   But empty cars still had to somehow get back 
to their owners.  So such cars were carried "for free" but lower in 
priority.  Similarly, perishable products (such as fruit) got express 
service and special handling such as refrigeration. Fruit spoils much 
more rapidly than coal.  Latency was critical then.  Still is today.

Such thinking was a motivation behind the inclusion of "Type of Service" 
in TCP4.  But today it seems that the only metric for network service on 
the Internet is bandwidth.   A 100-car train of coal can move a lot of 
weight over a long distance.  But a "Pacific Fruit Express" train could 
get those California fruits to the markets of the East Coast before they 
devolved into garbage.   The Internet today can readily handle my backup 
needs.   But it can't always deliver my interactive video.

Networks since ancient times have also experienced "dropped packets".   
Right now, archaeologists are exploring newly found parts of Pompeii.   
I suspect they'll find amphoras lost centuries ago.  Similarly, cranes 
are working in Baltimore harbor to retrieve containers that fell into 
the ocean just a few weeks ago - more "lost packets".

I still think that there's a lot to be learned by looking at the history 
of such "transport infrastructures" to see how and why they chose 
certain techniques and mechanisms, what the experience was in real use, 
and how the lessons learned can be applied to the Internet infrastructure.

Jack Haverty


On 4/15/24 11:28, John Levine via Internet-history wrote:
> Back in the 19th century there were a lot of railroads built in a lot
> of incompatible ways. The most obvious incompatibility was track gauge
> but there were others including the couplers between the cars and the
> ways they did (or sometimes did not) ensure that there was only one
> train at a time on each piece of track.
>
> These days most of the world has converged on standard gauge but there
> are still places like Spain and Russia that use broader gauges, and
> mountain railways and trams that use narrower. When a passenger or
> freight train crosses a border there's a variety of approaches, some
> of which may seems kind of familiar.
>
> The conceptually simplest approach is a gateway, at the border
> everyone gets off one train and gets on another.  The Canfranc
> station in the Pyrenees at the France-Spain border was famous
> for this.
>
> https://en.wikipedia.org/wiki/Canfranc_International_railway_station
>
> Another approach is layering. At the border, equipment lifts the car
> bodies off the bogies of the old gauge and puts them onto bogies of
> the new gauge. This is better since passengers don't have to get out
> (often from sleepers in the middle of the night) and goods don't have
> to be unloaded. This technique was patented in 1876.
>
> Here's the Prague-Moscow train changing gauge in Brest, Belarus.
>
> https://youtu.be/2nI467sc-Eo?si=w783HVwUGXAmQD7_
>
> Yet another approach is parallel operation, dual or triple gauge, with
> three or more rails allowing trains of different gauge to run on the
> same route. In Japan the Shinkansen are standard gauge but older
> railways are mostly 1067mm so there's a fair amount of dual gauge in
> and out of cities.
>
> This is a very old solution. The Niagara Falls bridge in 1855 had four
> rails for three different gauges, although now it's down to two.
>
> Here's a video of a dual gauge Shinkansen route:
>
> https://youtu.be/0d0XAaqEZ0s?si=ZYo27gNoAAXibtVq
>
> Another approach is switching on the fly. Some trains have variable
> bogies that can change gauge as the train is moving, which is pretty
> cool.
>
> Here is a Swiss train doing that:
>
> https://youtu.be/H0gj2LWe-SI?si=7zpZFc-jQPTIv8TX
>
> And a tutorial in Spanish:
>
> https://youtu.be/y8N7Ikw87tM?si=bCwrx5ph3SpgrevM
>
> The last approach is a flag day. One of the reasons the south lost the
> US Civil War was that they had a fragmented rail network, which
> continued to inhibit recovery and development after the war. So over
> two days, May 31-Jun 1, 1886, southern railroads regauged 11,500 miles
> of track to the Pennsylvania's gauge (1/2" wider than standard but
> close enough) and changed the bogies on the rolling stock.
>
> Here's a video about it:
>
> https://youtu.be/4v81Gwu6BTE?si=Yi9JDSU0onABpWju
>
> R's,
> John

-------------- 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/20240415/9e36230d/attachment.asc>