Re: [apps-discuss] Concise Binary Object Representation (CBOR)

[Phillip Hallam-Baker <[e-mail]>]

On Fri, May 24, 2013 at 7:11 PM, Cullen Jennings (fluffy)
<[e-mail]>wrote:

>
> Phil,
>
> I think your email is out of line and that behavior like this is the
> largest threat to the IETF begin relevant in the future. If we treat people
> like this, they are going to take their work elsewhere. I've talked to
> people about why they don't bring standards work to the IETF and the list
> of reasons why is surpassingly short and pretty well illustrated by this
> thread.
>
> Cullen
>

I was responding to Paul Hoffman. I hardly think he is quite the shrinking
violet you suggest or a novice likely to be deterred from IETF involvement.
He certainly isn't going to be deterred by anything I say.

In particular I was responding to his rather premature claim that the
difference between his proposal and the others was IETF 'consensus'. We
have had all of two days discussion of this proposal while the MongoDB folk
have a whole community built around theirs.


More generally, one of the reasons I see people not wanting to bring work
to IETF is the not invented here attitude. I was more than a little annoyed
when the IETF response to SOAP was to quick march BEEP to proposed standard
in an attempt to trump W3C and OASIS and so were many others.


On May 23, 2013, at 2:14 PM, Phillip Hallam-Baker <[e-mail]> wrote:
>
> > On Thu, May 23, 2013 at 3:24 PM, Paul Hoffman <[e-mail]>
> wrote:
> > On May 23, 2013, at 11:35 AM, James M Snell <[e-mail]> wrote:
> >
> > > That's well and good, from everything I've seen so far in this thread,
> > > the collective majority opinion can be summarized as "Ugh... Groan..
> > > Another one? Really?"
> >
> > Just a note that this is one of the only ones that people are groaning
> about that has an Internet Draft and might go through the IETF consensus
> process. Carsten and I (maybe naively) thought that doing this in this
> environment, instead of say posting ephemeral specs on a web page and not
> having it be clear where the community fit it, was a good thing.
> >
> > I don't remember you being appointed to address the issue.
>
> Actually, I think you, and the rest of the IETF did appoint them. We
> encourage people to bring good technical work and discuss it. Clearly
> binary encodings is an important topic for IETF protocols - This seems like
> a perfectly reasonable list to bring the email discussion to.


I did not object to it being brought to the list. What I objected to was
the claim of IETF consensus as the distinctive advantage for the protocol.

The IETF does have a body that is officially tasked with such things
although it doesn't actually do that work. It is called the IAB. Paul's
argument would make sense if it was being made by a member of the IAB
tasked with driving us towards consensus on a binary encoding. Absent an
explicit appointment to address the issue Paul's proposal is just another
Internet Draft.


What I interpreted the majority of the comments preceding it to be saying
was 'why not document one of the existing specs' which is of course what
was done with JSON itself. Instead we have a completely new proposal with
no deployed base being proposed and the only advantage to this proposal is
that it has an Internet Draft.

If the only problem with BSON etc is that they are ephemeral then making
them concrete might be a good start. If there are other problems then a
description would be useful.



> > Since almost all the response to your proposal has been that people
> don't like your design choices, and since you make it abundantly clear that
> you are not interested in our input, I can't quite see where a consensus is
> going to come. Unless that is the consensus is that your proposal sucks.
>
> That was not my read of the thread so far.


I saw several people raise requirements, all I saw was pushback of the form
'we have decided to do it our way'. I have not seen any of the design
choices being justified against requirements. I have not seen a technical
argument being made for any of the choices other than for the counted vs
delimited issue whare the answer was 'because'.


"Use an existing scheme" would be a perfectly reasonable proposal.

"Develop a new scheme based on requirements proposed by IETF community and
those in the BSON etc community" would also be a reasonable proposal.

"Use our scheme, it has an Internet Draft" does not seem to be a reasonable
proposal.



> > I would certainly object to a format that was counted being granted any
> degree of IETF recognition lest someone try to force me to use it in the
> future.
>
> This is not BCP that must be used by all future IETF protocols. It's an
> option for work that gets consensus to use it.
>

I don't believe that for two reasons.

First I have heard that exact statement used before about many similar
projects and then I have sat in rooms where the people who said that use
would not be mandated are telling another group that they are expected to
use it. Remember when people were being told that they had to use BEEP for
their Web service? I do.

Second, this particular spec is like JSON in that it only has value if
everyone uses the same thing. There really is not room for more than one
binary encoding of JSON in IETF.


JSON was originally an informational RFC. If we are going to have a binary
encoding then we should probably follow the same approach and that
candidates should be published as informational and we should let the
community make a consensus choice from amongst them.

I see now that my real objection is to CBOR being initially standards
track. What we should do is to begin with Informational descriptions for
existing standards that are widely used (MongoDB's BSON for example) and
Experimental proposals for anything new. Then we should see if there is the
possibility of convergence on one of the proposals.


I dropped a note into the BSON mailing list to see what their response
might be but with no reply so far.


-- 
Website: http://hallambaker.com/