I've read some BGP Limitations in student guide of BGP.
"BGP does not enable one AS to send traffic to a neighbor AS intending that the traffic take a different route from that taken by traffic originating in the neighbor AS."
Is it 100% correct? As i assume that we can control this behavior by various policies, at least when we are peering with provider (where provider is in touch with us).
Any other limitations in theory or practice?
Yes it is true.
You can certainly try to influence path selections by using policies but not force the other AS unless he is willing to accept your suggestions
Eg.. your ISP is connected to 2 ISPs A & B
if this ISP by default sends all the traffic to upstream ISP A for internet you cannot force your ISP to take ISP B route for your internet traffic
That sentence that you quoted is a strange sentence indeed.
Yes, you are right in a way. If we are best friends with our peers, I suppose we can try to convince them to do whatever we like. But sometimes we might have conflicting interests. For example, you might want your traffic via a particular low-delay link, but they won't let you go that way, because they are trying to balance their client traffic over their various available paths. This type of conflicting interests makes hard to control your traffic after it exits from your AS and before it enters into your AS.
And here is another thing. Most clients of a provider think they are the best client of the provider ever. For this reason, providers conclude on their policies (what they can do for customers and what they won't). The provider might be in touch with you, but the people who enforce the routing policies hardly ever directly talk with clients. So, the sales representative might be telling you that anything you want will be done. However, what is best for the provider network is in many cases the job of somebody else.
One last thing to add. Per client routing policies are not scalable for the provider to maintain and are normally avoided, especially if they involve a lot of (hard to remember) tricks to accomplish the desired effects.
I've seen your posts in the last couple of days. Though lengthy but very nicely explained. Then I had to dig your profile to go thru some other posts that may be of my interest. But was suprised that you are not regular on this forum but have been participating since 2005.
The one that i really liked from your archives was the reply to Kevin's posting on OSPF down state. I can just say Fantabulous. Also was amazed to know, there are people so crazy in this field who could go thru RFC's cover to cover.
You people are great inspiration for we newcomers. Wish you could be regular on this forum as other seniors & help people learn.
Hello to you (cannot find out your first name in your profile :-),
I sometimes feel bad when my posts get lengthy. I am not doing it on purpose. Although I consider good explaining something fully, in some cases there should be less lengthy ways to achieve the same effect. So, I am amazed myself to know that some people are crazy enough to read my posts from beginning to end, especially the one you mention, which I do not remember well, but I do remember that it was long enough for me not having the courage to re-read it now :-)
I have not been regular for various reasons, the primary reason being my school obligations for my master, which are still not complete, but yet close to completion.
See you people on the network and perhaps some of you in Networkers 2008 play-school in Barcelona.
Happy New Year,
We might see each other in Barcelona then! In fact, it would be great if some of the NetPro regulars could meet up there. What do you think? I know that Russ White is going, bacause he is on the speakers list. Have you been to previous Networkers?
Oh, I would not miss Russ White's session for the world! I registered early and sure thing I will be there. This is my first Networkers and I just can't wait to be there! Let's open an individual conversation for this NetPro meeting in Networkers "issue", because we have sort of hi-jacked this thread :-)
About the sentence that you quoted, now that I read it again I think the poet meant to say that the neighboring AS does not look into the source IP address (no policy routing based on source address) while forwarding your OUTGOING traffic. Let's say you are sending traffic towards destination www.cisco.com. Destination is www.cisco.com, not your own networks that you announce to the world via BGP. Somebody else announces to the world via BGP the route to www.cisco.com. Your outgoing traffic towards www.cisco.com will, in general, follow a path via the neighboring AS no different than any traffic that originated in the neighboring AS. They might be using some per-source/destination CEF load-balancing, but I would not consider this much of a consistently used different path, this is more of a random choice (and still not done via BGP). Perhaps your traffic might in the end use a different path than "some other" traffic that originated in the neighboring AS, because some exit point towards www.cisco.com in the neighboring AS is closer to the router you peer with, but then again, YOU cannot use BGP to enforce this. Even in this case, traffic originated in the neighboring AS that is "close" to the same router, will try to use the same "close" exit point.