W3C home > Mailing lists > Public > ietf-http-wg-old@w3.org > September to December 1995

Re: Proxies and loops

From: Simon Spero <ses@tipper.oit.unc.edu>
Date: Mon, 27 Nov 1995 12:18:20 -0800 (PST)
To: Jeffrey Mogul <mogul@pa.dec.com>
Cc: Luigi Rizzo <luigi@labinfo.iet.unipi.it>, http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com
Message-Id: <Pine.SOL.3.91.951127114736.25364E-100000@chivalry>

There are also reasons why the Forwarded header can lead to incorrect 
behaviour in a network of proxies interlinked by long-lived connections.
There are plenty of cases where forwarding a query on through a proxy 
though which it has already passed is in fact correct behaviour. 

For example, Suppose we have a network of servers in the following topology


    1    2    3    4    5
  A----B----C----D----E---F		
   \		    	 /
    \-----G-------------/
      6        7


Now, lets assume we want to sent a query from A to E. The network 
topology tells us that we should use link 6 and send the query via G and F

Suppose however, that in the meantime  F crashes, bringing down links 5 
and 7.  When the query arrives at G, the routing tables will have changed 
to make the correct next hop A again (A will have updated its tables to 
route queries to E via B).

With a Forwarded header, A would be forced to discard the message, even 
though a path exists between A and E. 
Received on Monday, 27 November 1995 12:21:51 EST

This archive was generated by hypermail pre-2.1.9 : Wednesday, 24 September 2003 06:31:36 EDT