Re: KHALED Routing Protocol (KRP).

[Lee Howard <lee@xxxxxxxxxx>]

From: ietf <ietf-bounces@xxxxxxxx> on behalf of Khaled Omar <eng.khaled.omar@xxxxxxxxxxx>
Date: Friday, April 28, 2017 at 8:08 AM
To: ietf <ietf@xxxxxxxx>
Subject: Fw: KHALED Routing Protocol (KRP).

Regarding the difference between tables, KRP within each region uses a table of only RBRs and RRs, the table is not listing prefixes because KRP uses the information stored on every single IP address and the IGP routing table, also, the Forwarding table details of one region is not advertised to other regions because they don't need to know it, enough to know how to reach the RBR and then it can deliver the packet faster and correctly.

Sorry, you’re carrying full routes for every /32 and every /128 in your IGP?

KRP will not require renumbering (reallocation) of IP addresses after the modification made to the 1st version of the draft that used the 1st two groups of IPv6 addresses as the KRP ASN, but the 2nd version used the middle two groups, and that will make the usage of KRP easier.

What about IPv4?

Finally, don't stop on what you think is not applicable, share your thoughts to add a solution to make things applicable thats the aim of the discussion, because KRP can route packets successfully from the source to the destination using the Best best route and with the lowest table entries for faster routing decisions.

Define “Best best.”

Shortest AS hops? Fewest intermediate devices? Cheapest link? Least saturated link? Hot potato? Lowest packet one-way latency?

I agree with Dave; the 6renum WG published rfc7010 IPv6 Site Renumbering Gap Analysis and then shut down; filling those gaps would be useful.

Lee

-------- Original Message --------
Subject: Re: KHALED Routing Protocol (KRP).
From: Dave Cridland
To: Khaled Omar
CC: ietf@xxxxxxxx

On 26 April 2017 at 00:27, Khaled Omar wrote:
> https://datatracker.ietf.org/doc/draft-omar-krp/

I like this idea, but I think it's sufficiently similar to the
original aim of EGP/BGP that it's already been proven impractical, and
moving to it would also be impractical. I am not a routing expert, so
I'll cheerfully wait being shot down by others, but this is my
rationale:

The reason why we have large routing tables within BGP routers is
mostly that ASNs now consist of many more prefixes than were
originally expected, and so the prefixes do not aggregate well enough.

This proposal involves, essentially, renumbering the entire internet
(which is very difficult in practical terms), such that aggregation
then works more efficiently. However, history has shown that while
this would work for the short term, it would require continuous
renumbering in response to many factors (new addresses, corporate
mergers, etc) in order to maintain its efficiency.

Since we have not solved renumbering very effectively in the first
place, I feel that this is not worth working on.

I would, however, suggest that addressing renumbering would be an
excellent use of your time. A viable solution to renumbering large
public networks would mean that the BGP table would shrink, by the
same logic as above.

That is, if we had a method for renumbering continuously, then it
would improve the efficiency of BGP as well, to almost the same
extent, without incurring the initial internet-wide renumbering
effort.

Finally, I would note that while routing tables are large, the
internet still functions - this may not be the best use of your time
therefore.

Dave.