oops, missed your observation re SACK. Yes, that firewall would get pretty well 
screwed by an encrypted SACK option.

> On Apr 29, 2016, at 11:29 AM, Scharf, Michael (Nokia - DE) 
> <[email protected]> wrote:
> 
>>> On Apr 28, 2016, at 9:26 AM, [email protected] wrote:
>>> 
>>> I guarantee someone will show us a middlebox that NEEDS to modify
>> every
>>> option we currently have and every option we will ever create.
>> 
>> I'd put it the other way around. The fact that middleware *can* get to
>> something lets someone find a use for doing so. That's not an argument
>> for it, it's an observation of fact.
>> 
>> Frankly, the only option I'm aware of people actively modifying is the
>> MSS. In a world in which ICMP is filtered and RFC 4821 isn't widely
>> implemented, having a VPN Tunnel Endpoint or other middleware reduce
>> the MSS to some value is a PMTU solution that mostly works. My
>> preference would be RFC 4821 or something like it. But that's not where
>> we are.
> 
> In TCPM there have been reports about sequence-number randomizing firewalls:
> 
> https://www.ietf.org/proceedings/87/slides/slides-87-tcpm-11.pdf
> 
> From slide 2: 'Firewalls "fixed" it by randomizing TCP sequence numbers - But 
> some of them forgot SACK'
> 
> Encryption the SACK option would result in the same issues like this entirely 
> broken firewall.
> 
> Yet, this shows that there may be cases in which *not* modifying the SACK 
> option creates problems.
> 
> (I'll obviously not argue in favor of this sort of firewall, as a modern 
> high-end TCP stack should have sufficient sequence number randomization.)
> 
> Michael

Attachment: signature.asc
Description: Message signed with OpenPGP using GPGMail

_______________________________________________
Tcpinc mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/tcpinc

Reply via email to