Re: Is Fragmentation at IP layer even needed ?

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Mon, Feb 8, 2016 at 2:39 PM, Carsten Bormann <cabo@xxxxxxx> wrote:
Ronald Bonica wrote:
> The words "many" and "some" don't do justice to the conversation.  https://tools.ietf.org/html/draft-ietf-v6ops-ipv6-ehs-in-real-world-02 provides more concrete numbers from real-world observation.

Ah, but the result is much simpler.

Some other real world data (Google QUIC experiments) already tell us
that a sizable part of the Internet (was it 7 %?) is not reachable via
UDP at all. 

I think you may have misunderstood Jana's numbers slightly.  The graph he showed in the Bar BoF did show that QUIC was not used in 7% of the cases, but that's not the same as saying it was unreachable. 

Chrome races TCP and QUIC and if TCP wins, it won't try QUIC again for the domain for a set interval; it then exponentially backs off with subsequent attempts.  Note as well that QUIC's handshake packets are a fixed size (1392 bytes), and always the largest packet in the connection.  So you may be seeing some race loss, some path MTU issues and some blockage of UDP that is port specific.  This last can occur when firewalls do not expect UDP on that port and block it on that port, but permit it on other ports (e.g. in the range used by RTP).

I've added Jana, in case he should have updated information or corrections.

This just ups that number slightly for IPv6 and UDP
protocols that don't have their own segmentation.

UDP, it was nice to have known you.


It seems a bit premature to say good bye,

regards,

Ted
 
Grüße, Carsten



[Index of Archives]     [IETF Annoucements]     [IETF]     [IP Storage]     [Yosemite News]     [Linux SCTP]     [Linux Newbies]     [Fedora Users]