Hi Nitin,

Thanks for your prompt response!

I removed parts that we agree.

> >
> >If so, why not apply this to the <ipv4-prefix> and <ipv6-prefix> as
> >well ,hence <ipv4-prefix> ::= <IPV4_ADDRESS> <IPV4_ADDRESS_LENGTH>
> >would be <ipv4-prefix> ::= <IPv4> <IPV4_ADDRESS>
> <IPV4_ADDRESS_LENGTH>,
> >and <ipv6-prefix> ::= <IPV6_ADDRESS> <IPV6_ADDRESS_LENGTH> would be
> ><ipv6-prefix> ::= <IPv6> <IPV6_ADDRESS> <IPV6_ADDRESS_LENGTH>
> 
> I think I know what you mean. What might be more appropriate would be:
> 
> <ipv4-route> ::= <IPv4> (<destination-ipv4-address> | <source-ipv4-address> |
>                          (<destination-ipv4-address>
> <source-ipv4-address>))
> 
> Or the other way to model this would be:
> 
> 
> <match> ::= <route-type> (<ipv4-route> | <ipv6-route> | <mpls-route> |
>                           <mac-route> | <interface-route>) <route-type> ::=
> <IPV4> | <IPV6> | <MPLS> | <IEEE_MAC> | <INTERFACE>

I am OK with either way.

> >
> >
> >2. Section 6
> ><rib-family> ::= <IPV4_RIB_FAMILY> | <IPV6_RIB_FAMILY> |
> >                    <MPLS_RIB_FAMILY> | <IEEE_MAC_RIB_FAMILY>,
> >
> >and
> >
> ><match> ::= <ipv4-route> | <ipv6-route> | <mpls-route> |
> >               <mac-route> | <interface-route> There are four rib
> >families defined, but there are five types route, so which rib family
> >does the interface-route belong to? Seems there needs a
> ><INTERFACE_RIB_FAMILY>.
> 
> An interface is part of a routing-instance <routing-instance> ::=
> <INSTANCE_NAME>
>                        [<interface-list>] <rib-list> [<ROUTER_ID>]
> 
> The purpose of a RIB family is essentially to identify what protocol 
> treatment to
> give to a packet. For example, for an IPv4 packet, you would do a TTL check. 
> For
> an interface-route, there is no such behavior. An IPv4 route can be associated
> with any interface in the <interface-list> above.
> Interface-route is more of a container (match) for all packets coming in on an
> interface. The network-device can choose to do DPI on the packet and run
> checks before it processes interface-route matching packets, or it can just 
> choose
> to do what the interface-route says (match all packet coming in interface-A 
> and
> send them out via interface-B).
> 
> 
> So in summary, I don¹t see a real need for INTERFACE_RIB_FAMILY.

OK.

> 
> 
> >
> >3. Section 6
> ><route> ::= <match> <nexthop-list>
> >               [<route-attributes>]
> >               [<route-vendor-attributes>]
> >
> >When a route was installed in the RIB, there should be an indication to
> >identify from which protocols (e.g., ISIS, OSPF, BGP, I2RS, CLI etc.)
> >the route is. So there may need a <Origin> attribute.
> >Maybe similar like this:
> ><Origin> ::= <RIP> | <OSPF> | <ISIS> | <BGP> | <LDP> | <RSVP-TE> |
> ><CLI>
> >| <I2RS>
> >
> >How do you think?
> 
> Yes, there ³could² be such an optional attribute. What we need to think more
> towards is, is just the protocol-type sufficient or do you need other things 
> as well,
> like protocol peer address (just thinking aloud).


At present, protocol-type is the straightforward one, I am not sure the 
protocol peer address is needed in the rib, there needs a use case to justify 
it. Another attribute may be needed is the I2RS client identifier, since there 
will be multiple clients that may install route into the rib.

> 
> >
> >4. Section 6
> ><tunnel-type> ::= <IP> | <MPLS> | <GRE> | <VxLAN> | <NVGRE>
> >
> >Should the <IP> be more specific to <IPv4> and <IPv6>? What does the
> ><IP> really intend for? UDP, TCP? And there may be other tunnel, e.g., l2tp.
> 
> Makes sense to have IPv4 and IPv6 called out separately. The IPv4/v6 is 
> basically
> for IP-in-IP tunnels. And yes, there could be other tunnels like L2TP, etc. I 
> guess
> we don¹t want to create an exhaustive list of tunnels in the info model, but
> rather leave it to the data model and data-model extensions. We¹ll split the
> v4/v6 stuff in next rev.

OK.

> 
> 
> >
> >5. Section 7.1
> >
> >AS-data information and AS length comparison are BGP related stuff,
> >this should be transparent to RIB manager. I am not sure whether this
> >belongs to Rib info model.
> 
> This has been removed in the latest version.

OK.

Thanks
Mach

> 
> Thanks
> Nitin
> 

_______________________________________________
i2rs mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/i2rs

Reply via email to