below...

On 2011-05-07 07:50, George, Wes E [NTK] wrote:
> -----Original Message-----
> From: Thomas Narten [mailto:[email protected]] 
> Sent: Friday, May 06, 2011 9:27 AM
> Subject: Re: Flow label drafts updated
> 
> Is the UDP port number mutable? Is the TCP sequence number immutable?
> [WEG] I think both are immutable because there's a checksum to detect 
> changes. That doesn't make it impossible to change it in the
> middle, as you mention, but it makes it much less likely that it will be.  
> 
> There are ways of modifying them that are undetecable. Does that make them 
> mutable?
> [WEG] Would you write similar language into the TCP and UDP spec like this 
> that claims immutable but that implementers should assume
> that it'll be changed along the way? By my interpretation, it comes down to 
> level of risk for modification. IMO the checksum is a
> significant difference and means that there is a reasonable assumption that 
> things won't be changed along the way. At the very least
> it'll detect changes due to data corruption, which you can't say for the IPv6 
> header.
> 
> The intention is that the Flow Label not get modified. Doing so can impact 
> its usefulness. This is true of most fields in packets.
> If random actors start tweaking various fields in a packet, that tends to not 
> be helpful (or worse).
> [WEG] Absolutely agree, but I don't think that calling it immutable is the 
> only way to say it, especially if you bracket it with a
> qualifier that says it's likely to get changed anyway.
> 
> I do think the document needs to say something about covert channels and 
> border routers zeroing out the field (and not just stick
> its head in the sand and try to have it both ways, which the wording Wes 
> quoted effectively does).
> But that doesn't mean we are declaring that the field is "mutable", implying 
> that anyone can start doing with it what they want.
> [WEG] fair enough, but I think this is sort of like the line between MUST NOT 
> and SHOULD NOT.
> What you're saying here is you SHOULD NOT change the value, while 
> acknowledging that it's likely to happen in certain applications.

It's the first I've heard of it. You MUST NOT change addresses and port numbers
in a packet; not that RFC 791 and 793 bother to mention it, because it was
unimaginable in 1981 that anyone except an attacker would do such a thing.

I'm hard over on a MUST NOT. What we've been arguing with Thomas is really about
how to express the warning that the flow label might get undetectably changed by
an on-path device, even though that is against the standard. A node downstream
from the change can't tell that it was changed, whether maliciously or by a
misguided firewall.

   Brian


> To me, Immutable implies MUST NOT be changed, and while people may still 
> ignore MUSTs it's less likely.
> To me, "anyone can start doing what they want" implies a MAY be changed
> 
> Wes George
> 
> 
> 
> ------------------------------------------------------------------------
> 
> --------------------------------------------------------------------
> IETF IPv6 working group mailing list
> [email protected]
> Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
> --------------------------------------------------------------------
--------------------------------------------------------------------
IETF IPv6 working group mailing list
[email protected]
Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------

Reply via email to