Re: [Idr] IDR WG LC on draft-ietf-idr-ix-bgp-route-server (2/2/2015 - 2/16/2015)

[heasley <[e-mail]>]

Thu, Feb 12, 2015 at 03:10:23PM +0700, Randy Bush:
> think of it this way, some people use a gui to drive a large expensive
> blob of code which configures their routers in a [semi-]automated way.
> 
> others do not wish to develop and/or maintain that lorge blob of code.
> perhaps they don't have a large enough network to amortize the effort.
> so they use a web site (ixp-manager) to configure an ixp-provided route
> server.
> 
> my problem with the second case is that the control plane and data plane
> are not congruent (think datakit, atm, ...).  the draft is suggesting a
> hack to ameliorate my objection.

i understand your objection and solution, and agree that its more palatable
than a broken forwarding plane.  But, i still think this is all a bunch of
kinky stuff to solve a problem that isnt all that difficult to handle if
the RS doesnt exist.  ie: its just configuration, which is not difficult.
the automation of configuration can be handled separately and w/o the RS
or protocol changes; eg: the IXP could instead provide a webservice to
disseminate peer presence and supply a tool that participants run on their
management host to retrieve and netconf that to the devices.  the unix
approach - simple tools used together.