Andrew: BGP->Route object translator

Yes, I'd much prefer to have connectivity arranged for my customers
*before* we brought them up, but just to play Devil's Advocate, if providers
do this every other day and diff the new data against the last data, the 
new routes will eventually get there, and a lot of unrouted gunk won't 
(maybe we won't have half the routes in the dbase unrouted anymore).  If
they wind up sending everything every day, we might be able to use this
data to delete routes in the dbase that haven't been found in a long 
while too (a big window obviously - months).

It seems to me that once you've got a route in your table that isn't
heard elsewhere, you've got a problem.  But if you're telling your 
customers that they'll have connectivity to everywhere x days after
they're all set up, and the level of expectation changes, it might
just become a feature.

Selina

> I really think this is the wrong approach of generating your route
> entries from your routing tables. Any connection which is down will
> not be in the bgp routing tables. If Andrew or others really dont have
> a database of their routes. Then they can grab all the cisco configs and
> depending on whether they redist static or not grep all the network
> lines or ip route lines. Grabbing a snap-shot of a BGP RiB is bound to
> miss things.
> 
> 		--Tony.
> 
>  Curtis Villamizar <curtis at ans.net> writes:
>   * 
>   * In message <199505032118.AA26055 at cat.isi.edu>, Cengiz Alaettinoglu writes:
>   * > 
>   * > Curtis Villamizar (curtis at ans.net) on May 3:
>   * > > > This needs to be refined slightly.  The "show ip bgp" output contains
>   * > > > an AS path.  The last entry in the AS path will be the home AS.
>   * > > > Andrew is putting in AS701 for every route he knows which would be
>   * >                                ************************
>   * > > > incorrect for most routes.  
>   * > 
>   * > I know nothing about cisco's show ip bgp, but I assume in
>   * > 	show ip bgp reg ^$
>   * > reg ^$ matches the as path, i.e. empty as paths, i.e. the routes that
>   * > he learned from IGP.
>   * > 
>   * > Is this right? In which case, he is the true home as for those routes.
>   * > 
>   * > Cengiz
>   * 
>   * 
>   * I missed that.  I guess Andrew had taken the AS into account.  This
>   * doesn't help his attached customers with their own AS.  Now all we
>   * need is the machinery in the shell program to avoid resubmitting the
>   * same thing every day and to withdraw the route (resubmit the route
>   * object as withdrawn) when components haven't been seen in a very long
>   * time (for some value of very long).  I think Steve agreed to let me
>   * write this and make it available since it shouldn't take long at all.
>   * I'll probably get to it before May 8.
>   * 
>   * Curtis



-------- Logged at  Thu May 4 18:45:00 MET DST 1995  ---------