Re: [dispatch] draft-devault-bare-07 to be discussed during IETF 114

[Eric Rescorla <[e-mail]>]

On Mon, Oct 9, 2023 at 5:24 AM Jiri Vlasak <[e-mail]> wrote:

> On Thu, Sep 28, 2023 at 02:54:33PM -0400, Dale R. Worley wrote:
> > I'm late to this party, but two things strike me immediately:
> >
> > Eric Rescorla <[e-mail]> writes:
> > > 1. The convention in IETF is to have things in "network" (big-endian)
> order.
> >
> > Yes; we've settled on consistently using big-endian order.  Is there a
> > reason for the IETF to bless an encoding scheme that is little-endian?
>
> BARE use-cases section includes "A structured data format for storing
> data in persistent storage" [1] (along more examples that invove network
> communication). The little endian is chosen because it is widely used in
> processor architectures.
>
> Is that a sufficient reason? If there is no chance that a protocol based
> on the little endian would be published in Standards Track, please, let
> me know.
>

I think you would need a very compelling reason, and "widely used in
processor architecture" isn't likely to do it. Note that gQUIC was
little-endian
and had significant existing deployment, but when the IETF standardized
QUIC, it made it big-endian.


> > Section 1 starts, "The purpose of the BARE message encoding, like
> > hundreds of others, is to encode application messages."  This is very
> > true.  But the I-D doesn't address why this encoding is decidedly better
> > than existing alternatives.  Presumably there's a case to be made for
> > it, but the case isn't made in the I-D.
>
> In general, it is better because of the BARE's goals:
>
> - Concise messages
> - A well-defined message schema
> - Broad compatibility with programming environments
> - Simplicity of implementation
>
> and how these goals are met.
>

To be honest, these all seem fairly subjective.

>From my perspective, the main question here is industry interest. There
are already a multiplicity of serialization formats, some of them with
very wide use. The IETF should not standardize yet another one
without strong evidence of interest. So far, I haven't heard any IETF
WG say they want to use BARE, nor have I heard any meaningful
demand from outside IETF. Absent that, I don't think we should
standardize it.

-Ekr