On Mon, 19 Aug 2002 [EMAIL PROTECTED] wrote:
>       i guess you are confused.
> 
> >But that doesn't mean that we break/refuse existing communications
> >earlier than we need to to achieve that.
> 
>       i have never said we should terminate existing connections.  i suggested
>       we should refuse new incoming connections (TCP SYN).

Note: 'communications' not 'connections'; the wording was deliberate, I 
think.

The point (as I saw it) was that if you want to deprecate an address, you
must first kill all references to that address.  If incoming connection
request comes in, probably either:

 1) the initiator has learned the address via some means (e.g. 
long-running application) some time ago before the address became 
deprecated
 2) the deprecated address is still referenced somewhere (e.g. DNS)

Sending RST might be ok for a purpose "ok, I'll reset this request
gracefully [compared to TCP timeout], round-robin or try some other
address", but for some others like "even though you don't know my other
address, I'll send a reset anyway" it may disrupt communications between
the two nodes.

We don't know the reasen the initiator wants to establish the connection,
so we probably should play safe and allow it.

-- 
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