Re: bettering open source involvement

[Alia Atlas <[e-mail]>]

On Thu, Jul 28, 2016 at 8:34 PM, Bjoern A. Zeeb <
[e-mail]> wrote:

> On 28 Jul 2016, at 21:06, Brian E Carpenter wrote:
>
> And there's our problem, right there. Protocols without APIs are
>> pretty much useless these days. IPv6 without a socket API would have
>> been an abject failure. Without RFC 2133, RFC 2292 and their successors,
>> who knows how the POSIX and Winsock support for IPv6 would have turned
>> out?
>>
>
> 2367 is a sad story for me personally as people went off and extended it
> everywhere and now we have an incompatible mish-mash in the world.  Just
> saying, maintenance is also important, and an easy and sensible way to get
> updates folded back in.
>
> The longer I think about publishing and obsoleting RFCs the more I want an
> Open Source model for them and put them in version control and just update
> them in place (not major extensions, but ..)—but that’s a different
> discussion.
>

How would that work combined with very wide deployment and independent
implementations?
I understand and sympathize greatly with interest in experimenting with new
methods - but I think it needs
more of a story that understands the stability assumed and doesn't cause
folks to have to revisit the
documents every year or less to see if their implementations need updating.

This is going to be an interesting situation with YANG models, where we may
see fairly rapid evolution.

What could we do different that still considers where standards are on the
innovation/stability curve?

Incidentally, I'm not sure where prejudice against APIs has come from.   Of
course, it doesn't frequently
come up in routing...

Regards,
Alia



> …
>
>>
>> but there are groups out there
>>> implementing IETF protocols and providing the APIs that allow
>>> application  developers to use those protocols and services.
>>> That is part of the open source landscape, as well.
>>>
>>
>> Sure. But if the protocol design, the API, and at least one implementation
>> aren't developed in lock-step, what on earth are we doing?
>>
>
> Writing RFCs which are 60ish pages long, use extra markers for the
> important bits and had no implementation after 5 years.  Can guess which
> one I was talking about?
>
> /bz
>
>