Re: [Spud] New Version Notification for draft-herbert-transports-over-udp-00.txt

[Michael Welzl <[e-mail]>]

> On 22. mai 2016, at 22.06, Ca By <[e-mail]> wrote:
> 
> 
> 
> On Sunday, May 22, 2016, Michael Welzl <[e-mail] <mailto:[e-mail]>> wrote:
> Hi,
> 
> I still didn’t get to read the draft - apologies!  I probably like it  :-)   but I hope it allows putting multiple TCP connections under one single UDP 5-tuple.
> Anyway, not speaking pro- or con- this draft, just a comment in line:
> 
> 
>> On 22. mai 2016, at 05.03, Jana Iyengar <[e-mail] <javascript:_e(%7B%7D,'cvml','[e-mail]');>> wrote:
>> 
>> Possibly unsurprisingly, I like this draft. I'd like to argue some of the choices made in the draft but I'll do that separately. For now, I just wanted to articulate a few thoughts as I caught up on this thread:
>> 
>> - The ship has sailed on new native-IP transports. I worked on SCTP for 6 years and tried and watched it get nowhere in terms of deployment over native-IP. It wasn't because it was in the kernel, it was primarily because of NATs and various middleboxes. Yes this is definitely my opinion, but arguing this point is silly. No new native IP transport has seen significant deployment over the internet over the past 20 years, and it's not for lack of trying.
>> 
> “The ship has sailed” and “it’s not for lack of trying” are both wrong IMO. How many applications have tried to use SCTP/IP and fallen back to TCP in case it doesn’t work? How many applications do you expect to change in order to get such a change to happen? And why should anyone expect someone designing or maintaining a router or middlebox to support native SCTP when they never even see the packets and never even hear anyone say “my application works better elsewhere because of SCTP” ?
> 
> This is what TAPS is trying to address. It’s a very deep architectural problem of the Internet - an inability to automate fall-backs, which unnecessarily burdens application developers. Rant, rant, rant, make me stop   :-)
> 
> Now all this being said, I also don’t fully get why some folks have such a big problem with running stuff over UDP. I’m not against doing that, if only as a temporary solution that would serve to convince people that the transport (option, ..) is useful.
> In a TAPS world, it’s just another option to try…
> 
> 
> The fact that udp is mostly (by volume) internet attack traffic  is my concern with udp. 
> 
> If legitimate traffic starts using udp in volume, it will make distiguishing and thwarting voulmetric attacks very difficult at scale. Without currently curbing n * 100g blasts of udp traffic with blanket policers, i would not be able to keep my network up... This is a daily issue.  For example, my usual udp volume is about 1%. If it goes to 10% suddenly, it is likely smart to drop 11% and onwards as a 10x increase in udp is 100% certainly not legit. 
> 
> My request to quic and spud is simply use a different transport protocol number so that their interesting and innovative traffic does not run up against the many network policers that are required to enforce well known baselines of good normal udp traffic. The quic folks say that they cannot do that since 10 year old cpe only passes udp and tcp, then they rant about ossified stacks and how we need to put everything on udp to make progess ... Seems like they are just choosing to ossify on udp ...  And udp is already considered trash (reflection attacks) ...So we agree to disagree, i guess 
> 
> https://tools.ietf.org/html/draft-byrne-opsec-udp-advisory-00 <https://tools.ietf.org/html/draft-byrne-opsec-udp-advisory-00>

Well - still wearing my TAPS glasses, I’d simply say “okay, in this case, try native too, and prefer that if it works".

Cheers,
Michael