[ih] Exterior Gateway Protocol

[Louis Mamakos]

On 2 Sep 2020, at 13:01, Noel Chiappa via Internet-history wrote:

>     > From: Grant Taylor
>
>     > Does anyone know of any surviving implementations of Exterior 
> Gateway
>     > Protocol, BGP's predecessor. ... Did any other network operating 
> system
>     > vendor or 3rd party vendor have EGP implementations?
>
> Liza Martin did one for the MIT so-called "C Gateway". That was 
> distributed to
> a number of people, along with the rest of the CGW code. All that code 
> is all
> still around (on the dump of the MIT-CSR machine which was recently
> retrieved), and easily available.
>
> I later re-wrote that one somewhat, to improve it, for the Proteon 
> router
> products. The one big change I recall which I made was to the code 
> which
> generated the lists of routes in the updates. To pack as many entriess 
> as
> possible into a single packet (IIRC, EGP routing table updates were 
> only
> single packets; no provision for overflow into sets of 2 or more), it 
> used a
> somewhat arcane organization, which a naive implementation would be 
> slow to
> generate. So I wrote code to walk the routing' table, and generate an
> intermediate tree structure which was a good match to the layout in 
> the EGP
> updates; the code to generate the output packets could then walk that
> swiftly. (Or, at least, that's my best memory; it's been ~40 years 
> since I
> last looked at it.)

As a practical operational matter, there was also concern about the ever 
growing
size of the EGP datagrams, which would get fragmented.  And now you're 
trying to
reassemble fragments, with ever decreasing probably of success in the 
ever-congested
Internet of the day.  And you have to wonder just how large a 
(fragmented) IP
datagram you could reasonably expect would be reassembled.

Of course, this was happening in the pre-CIDR times.

louie