Howdy,
You asked for community input, so I'll pipe up from the peanut gallery.
I happen to be looking for a JSON-ish on-the-wire encoding with binary support,
and I actually like what I see in CBOR. (I'll probably end up using
MessagePack anyway though... popularity has a quality all its own)
Comments inline...
On Aug 9, 2013, at 2:52 PM, Barry Leiba <barryleiba(_at_)computer(_dot_)org>
wrote:
To the rest of the community: Does anyone else think it is not
appropriate to publish CBOR as a Proposed Standard, and see who uses
it?
If the intention is to just document an encoding format/mechanism and see who
uses it, I think 'Informational' achieves that goal and is more appropriate.
If the intention is to base other standards-track work on this one in the
near-term, and you've got some such use-case in mind for it, then 'Proposed
Standard' seems reasonable to me if there're no strong objections. Afaict it's
always been easier to move Info to PS later, but it's been a lot harder to
deprecate PS. In the latter you have to prove something's broken/dangerous,
while in the former you just have to prove people use it.
Also, while I take Ted's point that people shouldn't feel constrained to use it
in other WGs, in practice I find that both WG participants and WG Chairs do in
fact give a huge bias to re-using something that's published as a PS. The
burden of proof required to not re-use an existing PS RFC is very high. They
say: "It's a STANDARD!" and the implication is its something written on a stone
tablet given to a guy named Moses.
I'm not saying that will happen in this case at all, but we shouldn't kid
ourselves that it doesn't matter. If it didn't matter, people wouldn't care
about labeling their IDs Informational or Experimental. People seem to *want*
the PS label, and I don't think it's because people want to upgrade to an IS
someday. [as an aside: that's what the 2-level RFC experiment should teach us -
that there is only 1 level that people care about, in practice]
To the rest of the community: What is your view of Phill's technical
arguments with CBOR? Do you agree that CBOR is flawed?
No, it doesn't appear to contain technical errors nor fail to meet its
self-stated design objectives. I think Phillip probably has different design
objectives. But I'm not an expert in the field of object encoding theory. :)
To the rest of the community: Do you agree with that concern? Do you
think such an analysis and selection of common goals, leading to one
(or perhaps two) new binary encodings being proposed is what we should
be doing? Or is it acceptable to have work such as CBOR proposed
without that analysis?
I think it's fair to publish CBOR as Informational right now. I think
publishing it as Proposed Standard would be ok if there wasn't strong
disagreement, but it appears there is strong disagreement, including on how it
came to be. I have no idea what the history of CBOR's development has been,
but if it's truly the case that it's just a couple people's personal work, then
that's cool but PS seems wrong to me if a third person says it's a bad design.
[again, I have no idea if that's true/false]
We rarely get 100% agreement even in WGs, but at least in WGs you get focused
concentration from many participants. Usually I've only seen PS going the A-D
route when there are no objections from anyone whatsoever. (but that could be a
wrong impression)
-hadriel