[ih] A paper

Tony Li tony.li at tony.li
Thu Jul 29 08:50:11 PDT 2021


>> Filder and I have published a paper recently about Internet protocols and
>> human rights but had a historical look at WHOIS, BGP/EGP and DNS. We
>> greatly enjoyed the informative conversation about BGP and EGP on this list
>> and helped us a lot with providing a more complete background.
>> 
>> Here is the link to the paper:
>> 
>> https://www.jstor.org/stable/10.5325/jinfopoli.11.2021.0376?refreqid=excelsior%3A5f6e0042f4bc042a36aa87e2a4d0107c#metadata_info_tab_contents
>> 
>> <
>> https://www.jstor.org/stable/10.5325/jinfopoli.11.2021.0376?refreqid=excelsior%3A5f6e0042f4bc042a36aa87e2a4d0107c#metadata_info_tab_contents
>>> 



Hi,

Thank you for your paper.  In it, you write:

"Ultimately, the intentions behind the design decisions, behind the design, and behind the impact of EGP and BGP are opaque.”

I think that there is definitely more that can be said.  It’s very clear that it was the intention of DARPA to get out of the business of funding and
running an operational network. Yet the architecture prior to EGP and BGP made that impossible. Technologically, it was also a clear
necessity to allow and control the interaction of multiple networks. It’s pretty clear that these dual motivations drove the original EGP design.
It’s equally clear that the technological design decisions of EGP (pure distance vector, fragmenting IP packets, lack of loop protection) 
made it impossible to scale. These issues drove the necessity for BGP.  

You also write:

"Cisco, a company formed due to the opportunities created by EGP and which employed one of the original BGP authors, 
portrayed EGP as a technical problem in interconnection efficiency.” 

It’s a little known fact that Cisco was not originally founded to build routers. The first intent was to build Ethernet hardware for DEC-10s and -20s.
It’s true that quickly morphed into ‘gateway’ development because after connecting a system to an Ethernet, it wasn’t going to be able to get 
past its local network. 

The issues with EGP, as described above, were not mere marketing bluster. They were fundamental limitations that made the network
unmanageable.  Network operators (at the time, NSFnet regionals) had to manually maintain lists of networks that were attached to other
autonomous systems and filter out EGP advertisements at each boundary as a means of loop prevention. This teetered on the brink of 
utter chaos for far too long.

Tony






More information about the Internet-history mailing list