[j-nsp] IS-IS now also results in ES-IS adjacencies?

[Hannes Gredler]

pekka,

Pekka Savola wrote:
> Hi,
> 
> It appears that sometime around 7.3 timeframe, the point-to-point 
> links between routers with IS-IS also seemed to start bringing up 
> ES-IS adjacencies as well (after IS-IS adjacency has been 
> established).

that is a result of moving ES-IS out of the kernel into RPD land.

> What's the deal?  Why enable it by default on all existing IS-IS 
> setups?  AFAICS, it's not useful between routers.

it was the "do not change the default-behaviour" mindset that
was guiding us here. the ever-since-behaviour was that as soon you had
a family iso configured on an IFL the kernel was generating
ES-IS ISHs. those ES-IS ISHs typically have been dropped at the
FPC ucode or at the kernel (no listener socket) level since we
do not need them

in 7.3 CLNS forwarding support has been added to the RPD/kernel/ukernel
here it is crucial that RPD does see ES-IS packets in case CLNS
routing is enabled and hosts want to tell their SNPA to RPD.
so thats the reason why you see the ES-IS adjacencies, now;
a result of carrying default-behaviour forward and adding a
ES-IS listener socket;

note that the ES-IS adjacencies to at no point affect the IS-IS module.

> Also, the docs seem 
> to be a bit of two minds about the usefulness of ES-IS between 
> routers:
> 
> "ES-IS must not be disabled."
> 
> "If ES-IS is explicitly configured and disabled, the interface does not 
> send or receive ES-IS packets."
> 
> (i.e., if ES-IS must never, under any circumstance be disabled, why is 
> there an option for disabling it?)

i guess it should read:

"for CLNS routing and direct connected ISO hosts, ES-IS must not be disabled."

tx,

/hannes