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

Re: [hybi] SPDY protocol from google frame



On 2009-11-13, at 9:49 AM, Roberto Peon wrote:

In the case if SPDY, we have a simple tool that reads the binary protocol and outputs ASCII in a correlated structured format.

We'll also have a wireshark plugin soon for "defense in depth." Debugging isn't significantly more difficult than it is for HTTP (where i have to look at the output in a hexdump format to see if the delimeters are correct, etc. The traffic generated by users is truly surprisingly weird sometimes)

There will be malformed SPDY traffic if multiple libraries
implement the protocol.  Will that be easy to debug with the
wireshark plugin?  Of course, debugging tools can be
developed for any binary protocol; the question is whether
you have that debugging tool available in the field when
you are faced with a critical bug.

And, thanks to the SHOULDS, you have no idea whatsoever how the next hop is going to construe the exact bytes that you've received. Will it be able ti handle a message with mixed \r\n and other eol terminals??

Perhaps the use of newlines in these cases should simply
be clarified?

Debugging HTTP on google servers is a royal pain. It isnt easy to debug just because it is text...

That's a controlled, high-volume environment, and I agree,
advanced tools are possible to install and necessary even
for debugging text protocols.

Would we implement IP or tcp in ASCII?

Only for the packets that travel via paper printouts ...

This is a good point; on the other hand, TCP is well
established, inherently fixed length (now that everybody
is using ipv6 we won't run out of space again), and
lower in the stack.

It's all a question of tradeoffs -- we could certainly
be successful with either a text or a binary protocol.

Ted.


On Nov 13, 2009 8:28 AM, "Ted Goddard" <ted.goddard at icesoft.com> wrote:


OK, let's say there are 1000 people debugging network
traffic at any time -- this is a dangerous subset of
people to ignore, though, because they're not just the
users of the protocol, they're the ones developing
the servers, clients, and applications for the users.

Ted.
On 2009-11-13, at 12:20 AM, Ian Fette (イアンフェッティ) wrote:

> Ok, I will give you a few thousand. Either way, it's many orders of magnitude fewer people as comp...

> _______________________________________________ > hybi mailing list > hybi at ietf.org > https://www....

_______________________________________________ hybi mailing list hybi at ietf.org https://www.ietf....





Note Well: Messages sent to this mailing list are the opinions of the senders and do not imply endorsement by the IETF.