[ih] Exterior Gateway Protocol

[Guy Almes]

Vint, Scott, et al.,
   You have both correctly remembered the uneasiness as being a vague 
sense of layer violation.
   Picking up on Scott's specific reason for the objection, recall that 
an Autonomous System was defined as a set of routers rather than (as in 
DECnet, for example) a set of network numbers.
   And so, even if the two "Border Gateways" engaging in BGP were in 
different Autonomous Systems, they typically shared a network number. 
For example, they might have two different interfaces on the same 
Ethernet or be at two different ends of the same serial line.

   So (if I recall correctly), it was not a layer violation, but the 
reason why was a little subtle.

   This all relates to the original thinking about Border Gateways. 
Later, people did start doing BGP other than between Border Gateways 
sharing a network number / prefix.  There was, of course, "Interior BGP" 
between Border Routers of the same AS.  And there was the later 
application of BGP to connect to route servers.
   Does anyone know whether any of those later applications did depend 
on inter-AS routing already working?  If so, how was the layer violation 
handled (or problems arising from it not being handled)?

	-- Guy

On 9/4/20 7:51 AM, vinton cerf wrote:
> It felt like a layer violation to run BGP over TCP but in fact it proved 
> to be very useful.
> 
> v
> 
> 
> On Fri, Sep 4, 2020 at 7:36 AM Scott Brim via Internet-history 
> <internet-history at elists.isoc.org 
> <mailto:internet-history at elists.isoc.org>> wrote:
> 
>     On Fri, Sep 4, 2020 at 1:29 AM Joseph Touch <touch at strayalpha.com
>     <mailto:touch at strayalpha.com>> wrote:
> 
>      > On Sep 2, 2020, at 2:33 PM, Guy Almes via Internet-history <
>      > internet-history at elists.isoc.org
>     <mailto:internet-history at elists.isoc.org>> wrote:
>      > ...
>      >  As Vint just reminded me, not only was the variable length AS
>     path an
>      > unusual characteristic, but BGP's use of TCP was quite
>     controversial. But,
>      > in addition to other advantages, the use of TCP made possible
>     several early
>      > implementations.
>      >
>      >
>      > Can you tell us more about what part of BGP’s use of TCP was
>     controversial
>      > at the time?
>      >
>      > I.e., other than the error of linking “this path is up” inside
>     BGP to “TCP
>      > is stable over that path”, a decision that required multiple fixes to
>      > address (MD5, TCP-AO, RST rejection, and route dampening).
>      >
> 
>     Layering. You're setting up a connection to a node in a different AS, so
>     you needed AS-level routing to tell you the path, so you need BGP so
>     etc.
>     -- 
>     Internet-history mailing list
>     Internet-history at elists.isoc.org
>     <mailto:Internet-history at elists.isoc.org>
>     https://elists.isoc.org/mailman/listinfo/internet-history
>     <https://urldefense.com/v3/__https://elists.isoc.org/mailman/listinfo/internet-history__;!!KwNVnqRv!TtUs5oq4r7wsE4CSc31u0IABrBoVSm1VmB55DmwmKFZe6ww41bLHZfm1sMyhAA$>
>