On Tue, 15 Jan 2002, Francis Dupont wrote:
>    In theory, yes.  But the question is, does it really make that much of a 
>    difference as long as at least one way is "unoptimized"?
>    
> => the same difference than when both ways are unoptimized or optimized.
> Difference in distance is the same, difference in byte overhead too.
> This is true for all points with one exception: protocol complexity
> because triangular routing has roughly the same complexity than
> bidirectional tunneling.

What about latency (e.g. interactive session) detected by the user? (might 
be the worst of the two)

Reliability is one argument too.

>    Practice?  I don't think there's much difference in bidir
>    tunnel/triangular routing; I doubt e.g. TCP would be able to get much
>    better performance from triangular than from bi-dir (and there are goals
>    like hiding your origins that triangular cannot solve).
>    
> => I can't see how TCP is involved. I believe your argument is different:
> if the traffic is very asymmetrical (more on one way than on the other),
> you prefer to have it on the optimized way...

I think I'd prefer to have it either _completely optimized_ or _completely
unoptimized_, not optimized to one direction and unoptimized to the other.

> I am afraid the real argument against the triangular routing is this
> discussion itself because the main triangular routing advantage comes
> from the (future) fact that HAO is supported by every IPv6 nodes.
> With enough FUD this shall never become true and the advantage shall vanish...

Yes.  But in all reality, a user who requires mobility (e.g. moving fast) 
would be a fool to rely that all CN's would implement this (or else the 
connection breaks) during the next 3 years or so..

Triangular routing relies on the existance of this feature everywhere.

Bidirectional tunneling always works, and in practise *may* (I don't know) 
be roughly equal in user-detected latency, TCP thoughput, etc.  Does it 
not seem Bidir is far superior to triangular routing?  (As a matter of 
fact, bidir does not really even require MIPv6 -- I don't think there's 
any feature missing in current specifications that would be required :-).

-- 
Pekka Savola                 "Tell me of difficulties surmounted,
Netcore Oy                   not those you stumble over and fall"
Systems. Networks. Security.  -- Robert Jordan: A Crown of Swords

--------------------------------------------------------------------
IETF IPng Working Group Mailing List
IPng Home Page:                      http://playground.sun.com/ipng
FTP archive:                      ftp://playground.sun.com/pub/ipng
Direct all administrative requests to [EMAIL PROTECTED]
--------------------------------------------------------------------

Reply via email to