Thanks for the comments. Extracting:

On 2010-04-15 22:14, Alexandru Petrescu wrote:
...
>> 1.  If and only if the flow label in an IPv6 packet has the default
>> value of zero, then a router MAY set it to a value between between 1
>>  and 0xFFFFF.  This option modifies the rule that the flow label must
>>  be delivered unchanged, by allowing exactly one router to set it if
>>  the source host did not set it.
> 
> I believe this may pose problems.  The case is when the sending Host1
> does not set the Flow Label and the receiving Host2 knows somehow (e.g.
> the app protocol is older than new Flow Labels) that H1 doesn't
> implement Flow Labels and so expects "0" zero from it.  When H2 sees a
> flow label set then it deduces something is wrong, because it knows H1
> couldn't set it.

Indeed, I think we should make it clear that receiving hosts MUST NOT
rely on the flow label for any kind of upper layer behaviour. That is
a missing rule in RFC3697, I think.

> In RPL there's talk proposing "restoration" of the Flow Label at the
> exit point, but I don't agree with that either.

That is *required* by the existing rules of RFC3697.

>> packets leaving the local domain may contain flow label values that
>> are not pseudo-random
> 
> For some reason I have difficulty understanding "not pseudo-radom": does
> it mean actually "random", interpreted as a double negation?  Does it
> mean "not random but can't talk random because true random doesn't
> exist"?  The implications of this clarification are wide I believe.

OK, good point. What is intended by "not pseudo-random" is that
it's either assigned systematically (e.g. 1, 2, 3...) or that the bits
are encoded with some special meanings. I don't like to write plain
"random" because most network nodes don't contain random number
devices.

> 
>> The flow label is not protected in any way and can be forged by an
>> on-path attacker.  On the other hand, a pseudo-random flow label
>> cannot be readily guessed by an off-path attacker.
> 
> That "OTOH" makes think that because pseudo-random flow labels cannot be
> guessed then the Flow Label is protected against on-path attackers.  I

What? Not at all. On-path attackers can observe and/or forge any flow
label. But as draft-blake-ipv6-flow-label-nonce shows, they can have
value for making off-path attacks harder.

> don't think much protection is afforded by the use of visible be them
> random numbers.  Were it public-private-keys then yes but here we don't
> talk Flow Labels as security keys, at least because of the obvious
> performance requirements.

Indeed not.

   Brian
--------------------------------------------------------------------
IETF IPv6 working group mailing list
[email protected]
Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------

Reply via email to