[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [RRG] getting rid of longest match
On Feb 25, 2008, at 3:06 PM, Iljitsch van Beijnum wrote:
On 25 feb 2008, at 23:43, Scott Brim wrote:
What I'm thinking is that for any given part of the address
space, there would be a fixed prefix length. So for 192.0.0.0/8 /
24, for 64.0.0.0/8 /20, for 17.0.0.0/8 /8... So in theory, there
would be no overlap. There could still be if the filters weren't
set up properly, which would probably have to resolved when
translating the RIB into the FIB.
Isn't this just what we did pre-CIDR?
Yes, except that it wouldn't be hardcoded into the IP stack but
enforced through filters. So it could be /24 for 193.0.0.0/9 and /
19 for 193.128.0.0/9.
The issue with overlap is: which one do you choose? Longest match
says the smallest block is the good one, but I don't think YouTube
appreciated that logic yesterday. If we ditch longest match we need
another rule. Allowing only one prefix size for a certain address
range is one way to resolve this, but others are also possible. For
instance, S-BGP or soBGP-like security mechanisms.
If I understand this proposal, then for any given prefix, the entire
net would agree on a prefix length and only propagate paths for this
prefix length. This would imply that there would actually be no
required hardware change, as all of this filtering could (and should)
happen during BGP processing (specifically during UPDATE parsing) and
prior to RIB insertion and thus prior to the FIB insertion.
Thus, from a hardware perspective, this is largely a NOP.
Note that this could somewhat decrease the complexities of future
hardware designs. Since we now guarantee that there are no
overlapping prefixes, the table now effectively forms a tree of
variable prefix lengths, but now the first match found in the tree
suffices. In the grand scheme of things I would judge this not to
have a significant impact, but it is arguably simpler.
Tony
--
to unsubscribe send a message to rrg-request@psg.com with the
word 'unsubscribe' in a single line as the message text body.
archive: <http://psg.com/lists/rrg/> & ftp://psg.com/pub/lists/rrg