Bruce M Simpson <[EMAIL PROTECTED]> wrote: > On Thu, Mar 16, 2006 at 07:35:20PM +0100, Bart Van Kerckhove wrote: >> Is this by design, or just lack of time/interest? >> If anyone feels up to the task of fixing/implementing what's needed >> to make this work, we'd be happy to sponsor its development. > > This is a collision between the connected route implicitly created > by configuring an interface which has the same prefix length as > an existing route in the FreeBSD FIB. > > This is a known issue and is by design. > > Most BSD-derived implementations have this limitation. It needs to be > resolved in preparation for equal-cost multipath. > ECMP was indeed one of the features i was looking for at that time, which i found to be impossible. I just don't like the idea of moving towards another platform just for this reason, since I'm very happy with freebsd's performance. There used to be a patch for ECMP, but it was a huge hack. very dirty at best.
> Sadly whilst I'd be more than happy to work on this (with or without > funding), I don't have the free time to do so, but I may be able to > eke out spare time to look at patches. Any quick hacks you have in mind that could allow multiple copies of the same route / pfxlen in the freebsd routingtable? Doesn't even need to have metrics in kernel-level, the point of the whole thing (for now) would be to be able to insert an additional route with the same prefix, and delete the previousy set route *afterwards*. Which is what ospfd/zebra would do. Any insights / hacks / other dirty things are very appreciated. For all others following this thread: "we'd be happy to sponsor its development" still is in effect ;) Met vriendelijke groet / With kind regards, Bart Van Kerckhove [EMAIL PROTECTED] http://www.it-ss.be - "Solid Solutions for your IT needs" "There are 10 kinds of ppl; those who read binary and those who don't"
smime.p7s
Description: S/MIME cryptographic signature
