Re: [hybi] Masking only Payload/Extension Data

[Greg Wilkins <[e-mail]>]

I would like to note a procedural issue that has occurred many times
in this working group.

Things like masked framing have been added to the specification
without consensus (there was consensus on masking, but objections had
been raised regarding the masking before it was added).   We then have
to argue to have the feature removed and if consensus is not achieve
then that feature remains.    This in effect means that the status quo
is to add a non- consensus feature.

Surely this should have been done the other way around.  A broad
consensus should have been achieved in the WG before such a
significant change the the agreed framing was put into the draft.

regards




On 11 March 2011 07:56, Brian <[e-mail]> wrote:
> Just a quick updated tally of support/opposition:
>
> By my count, we now have eight voices in favor so far, including myself:
> Andy Green
> Ytaka Takeda
> Greg Wilkins
> Willy Tarreau
> Joel Martin
> Brian McKelvey
> Bruce Atherton
> Julian Reschke
>
> Three on record as not having a strong opinion one way or the other:
> Ian Fette
> John Endicott (Generally in favor, but not strongly so)
> John Tamplin (Generally opposed, but hasn't mentioned a strong
> opposition that I've seen)
>
>
> And two opposed:
> Adam Barth
> Patrick McManus
>
> If I've missed or misrepresented anyone, please let me know.
>
> I agree with Bruce in that I feel that pretty much everything that can
> be said on the topic at hand has been said.  Sal, what is the next
> step for the WG toward a declaration of consensus one way or the
> other?  Do we need a hum or formal straw poll, or will this thread
> suffice as such?  Is further discussion warranted?
>
>
> Brian
>
>
> On Thu, Mar 10, 2011 at 11:59 AM, David Endicott <[e-mail]> wrote:
>> Certainly not for a computer.  (But I'm getting older and it's
>> becoming difficult to xor in my head these days.)
>>
>> But any extra work is extra work, and if the agent is manipulating
>> frames without processing them then its likely performance is a
>> concern.  Not that I really think xor'ing a few bytes is going to cost
>> enough clock cycles to matter.
>>
>> My vote is thus: ambivalent; with a logically weak prejudice for
>> unmasked headers.
>>
>>
>> On Thu, Mar 10, 2011 at 2:49 PM, John Tamplin <[e-mail]> wrote:
>>> On Thu, Mar 10, 2011 at 2:36 PM, David Endicott <[e-mail]> wrote:
>>>>
>>>> Afterthought:  an unmasked header would allow Websocket aware
>>>> intermediaries to manipulate frames without unmasking them.
>>>> Websocket aware load balancers and distributed frameworks come to
>>>> mind.
>>>
>>> Right, that has been the main objection (by my estimation anyway) to masking
>>> the header.  However, unmasking 2-10 bytes doesn't seem like a large
>>> burden.
>>> --
>>> John A. Tamplin
>>> Software Engineer (GWT), Google
>>>
>> _______________________________________________
>> hybi mailing list
>> [e-mail]
>> https://www.ietf.org/mailman/listinfo/hybi
>>
> _______________________________________________
> hybi mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/hybi
>