>Why isn't BGP a distance vector protocol?
If I were pushed on the subject, I'd probably call path vector, BGP's
algorithm, a superset of distance vector with respect to loop
detection.
To go more deeply into loop detection,
--1st generation DV (XNS RIP, IP RIP, RTMP): Loop avoidance through
split horizon and holddown. Loop detection through count-to-infinity.
See original Bellman-Ford papers and the RIP RFC.
--2nd generation DV (at least IGRP, not sure about IPX RIP): Loop
avoidance through split horizon and holddown. Loop detection through
count-to-infinity, and also by detecting pattern of steadily
increasing hop count or metric among successive routing updates. See
Cisco IGRP patent and various informal Cisco papers--I seem to recall
variously by Charles Hedrick, Dino Farinacci, etc.
--3rd generation DV/DUAL: Loop avoidance principally from reliable
update (i.e., no old or bad information), and following rules
including that a route with lower cost than your current one cannot
form a loop (great simplifications). Lots of theoretical
justification published by JJ Garcia-Luna-Aceves.
--Path vector. No concept of cost in the basic routing algorithm.
Does maintain a traceroute-like path vector indicating how the route
was created and propagated. Routes that contain your own AS, by
definition, show looping behavior and are rejected. Incidentally,
path vector is demonstrably loop-free only when unconstrained by
policies, but that issue gets VERY theoretical. I'd have to dig for
specific references, but I seem to recall Sally Floyd did some work
with AT&T researchers, and there are probably presentations at NANOG
and ACM SIGCOMM.
>Isn't the definition of a distance vector protocol in short- a
>routing protocol that determines the distance to all destinations
>from itself using information received from it's neighbors?
>Well the big difference I can see with BGP is that it uses a
>different way of figuring the distance- being AS's in it's path
>instead of hops or cost.
Huge difference in loop detection. Classical DV uses a sum of
interface costs, whether these are constant at 1 (i.e., hop count),
delay based (e.g., IPX RIP), etc. The route metric is a quantitative
sum of costs.
Analysis of the path vector is qualitative rather than quantitative.
In classical path vector, a route is either acceptable or not.
Unacceptable routes contain your AS number.
You may be thinking of AS path length as equivalent to an IGP metric.
Trust me without getting very mathematical or pointing to very
mathematical reference. It ain't. BGP just doesn't have a
sum-of-costs concepts. MEDs and local preferences are not end-to-end
with respect to the destination; they affect next hop decisions.
>If that were the thing that made it different, IPX RIP and DECnet
>Phase 3 wouldn't be distance vector protocols either, since they
>don't just use "hops" as IP RIP does. The reason I bring this up is
>that I have heard it called a path vector protocol. It just seems
>like marketing hype since Distance Vector protocols seem to have a
>bad connotation for the reason it is referred to as a path vector
>protocol.
Path vector has a very specific meaning. It's not marketing hype.
>What would you call IPX RIP then since it uses hops and cost.
Distance vector. Using incremental or periodic update has nothing to
do with a protocol being DV, LS, or PV. Summing costs, whether the
individual costs are hops, ticks, etc., is still a route metric.
>
>DECnet phase 3 uses incremental updates.
>
>>>>Brian
>
>>From: "Howard C. Berkowitz" <[EMAIL PROTECTED]>
>>Reply-To: "Howard C. Berkowitz" <[EMAIL PROTECTED]>
>>To: [EMAIL PROTECTED]
>>Subject: Re: Hybrid Routing Protocol
>>Date: Thu, 1 Feb 2001 14:07:30 -0500
>>
>>>The only DV protocol that sends incremental updates? You sure about that?
>>
>>For IP interior routing on Cisco, yes. If one wants to argue BGP is a
>>DV derivative, it sends incremental updates.
>>
>>Other DV incremental update protocols are described in routing
>>research literature, but they are not widely deployed.
>>
>>>
>>>>>>Brian
>>>
>>>
>>>>From: John Neiberger <[EMAIL PROTECTED]>
>>>>Reply-To: John Neiberger <[EMAIL PROTECTED]>
>>>>To: Jason Wong <[EMAIL PROTECTED]>, [EMAIL PROTECTED]
>>>>Subject: Re: Hybrid Routing Protocol
>>>>Date: Thu, 1 Feb 2001 10:20:42 -0800 (PST)
>>>>
>>>>Well, it certainly is called a hybrid, but that's marketing hype; it's
>>>>operation is completely DV in nature. It's "hybrid" characteristic is that
>>>>it only sends incremental updates and it establishes neighbor
>>>>relationships,
>>>>which other DV protocols do not do.
>>>>
>>>>That does not, however, change its basic nature, which is distance vector.
>>>>
>>>> > Actually EIGRP is a hybrid protocol. I believe it is the ONLY example
>>>>of
>>>>one, in fact.
>>>> >
>>>> > JW
>>>> >
>>>> > ------------
>>>> > A DV protocol, like RIP or EIGRP, send their entire routing table to
>>>>their
>>>> > directly attached neighbors and then receive their neighbors routing
>>>>tables
>>>> > in return. That's an important point: they send the *entire* routing
>>>>table,
>>>> > not just the routes they know about first hand.
>>>> >
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>_______________________________________________________
>>>>Send a cool gift with your E-Card
>>>>http://www.bluemountain.com/giftcenter/
>>>>
>>>>
>>>>_________________________________
>>>>FAQ, list archives, and subscription info:
>>>>http://www.groupstudy.com/list/cisco.html
>>>>Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
>>>
>>>_________________________________________________________________
>>>Get your FREE download of MSN Explorer at http://explorer.msn.com
>>>
>>>_________________________________
>>>FAQ, list archives, and subscription info:
>>>http://www.groupstudy.com/list/cisco.html
>>>Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
>>
>>_________________________________
>>FAQ, list archives, and subscription info:
>>http://www.groupstudy.com/list/cisco.html
>>Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
>
>_________________________________________________________________
>Get your FREE download of MSN Explorer at http://explorer.msn.com
_________________________________
FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html
Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
