Re: When is an idea a good idea?

Spencer Dawkins <spencerdawkins.ietf@xxxxxxxxx> · Wed, 29 Jan 2014 09:45:48 -0600

I'm replying to Brian and to Scott here ...

On 01/28/2014 05:11 PM, Brian E Carpenter wrote:
Hi Spencer,

I think there's a bit more to be said, though.

The goal is to make the Internet work better, not to make
private networks work better. Actually RFC 3935 defines
the IETF's scope as: "protocols and practices for which
secure and scalable implementations are expected to have
wide deployment and interoperation on the Internet, or to
form part of the infrastructure of the Internet."

So if there are two ways of doing something, one of which
works globally and one of which only works locally, the choice
is a no-brainer. If a local-only solution comes up to the IESG,

I couldn't agree with Brian and Scott more. I'm focusing on "comes up to 
the IESG".

My point is that the sooner someone asks "so, is this really intended 
for deployment on the global Internet?", the better. If working groups 
and authors thought about that before, or even during, WGLC time, I 
would like to think that we would have fewer cases where the question is 
asked for the first time in AD review, or in directorate reviews, or in 
IETF Last Call.

I don't think it's controversial at all that the IETF does better when 
documents don't encounter Late Surprises (tm). If the first time someone 
says "but this is only intended to be used when $whatever" is in an 
email exchange during IETF Last Call, that's a Late Surprise, and the 
chances that the best outcome will happen aren't going up at that point.

As I said in my note, this is my personal suggestion. I'm not asking for 
an Applicability Statement as a mandatory section in all IDs, precisely 
because of Brian's point that the default is "this proposal is suitable 
for deployment on the global Internet". I'm asking working groups and 
authors to consider whether that's true for their proposals, or if I 
should be reviewing with something else in mind.

And as always, I encourage people to Do The Right Thing.

Spencer

I think the first question is "Why isn't there a globally
applicable solution to this problem?". Then the second question
is "Why is this on the standards track, rather than being an
informational document for the record?". And the third question
is "Why is this better as an IETF document rather than an
Independent Submission?".

There may be perfectly good answers to those questions, but
IMHO they need to be asked.

Regards
    Brian

On 29/01/2014 09:14, Spencer Dawkins wrote:
Disclaimer - I'm one of the ADs who will be balloting your documents
until at least March 2015, but I'm only one of 15 ADs, and haven't
talked about this with the rest of the IESG.

The past week has brought me yet another example of an idea that could
be a good idea, but not under every possible set of conditions, being
evaluated as if we had to decide whether it was a good idea in all
cases, or a bad idea. This might sound familiar to you, but if it does,
let's not talk about that specific case.

The longer (and more painfully) we talk during Last Call, the more
details we unearth that help me to understand what document
authors/working groups are thinking, and that's good, but if it was less
painful, that would be even better.

http://tools.ietf.org/html/rfc2026#section-3.2 describes Applicability
Statements, as distinct from Technical Specifications, and says (among
other things):

    An Applicability Statement specifies how, and under what
    circumstances, one or more TSs may be applied to support a particular
    Internet capability.

...

    An AS may describe particular methods of using a TS in a restricted
    "domain of applicability", such as Internet routers, terminal
    servers, Internet systems that interface to Ethernets, or datagram-
    based database servers.

Speaking only for myself, I don't expect Proposed Standards to be
perfect, and to work perfectly in every situation, but if a
specification doesn't describe any limits on applicability, I'm going to
be evaluating it as if it will be used on the open Internet (that's what
the "I" in "IETF" stands for).

If a draft says it's only intended to be used within an IP subnet, I'd
evaluate it differently. I might ask that the authors/working group
consider what the TTL should be set to, so that what starts out within
an IP subnet stays within an IP subnet, but (to use one actual example)
we wouldn't be arguing about whether it's OK to send 32K max-length
packets over an arbitrary Internet path at line rate without getting any
feedback about path capacity - there are probably paths where that would
work, and there are definitely paths where it would not.

If a draft says it's only intended to be used on provisioned, managed
internetwork links subject to SLA monitoring, I'd evaluate it differently.

If a draft says "this is a hack, intended for use on old hardware that
can't do $X", I'd evaluate it differently.

There are other limited-use scenarios that would make me evaluate a
draft differently.

None of this is a guaranteed pass, but it would help me a lot. It would
likely help other reviewers, too.

So, if you don't intend for your draft to be used on the global
Internet, please say so! As per
http://tools.ietf.org/html/rfc2026#section-3.3, it's not necessary to
put an Applicability Statement in a different draft; just a section that
says (another actual example) "this has been tested using these
parameters on a lightly-loaded LAN, and it works there", that is more
helpful than a tug of war(*) about whether something is a good idea or a
bad idea in all situations, in front of a live studio audience.

Thanks to Alia Atlas for nudging me to think about this more.

Spencer

p.s. If you have feedback about what I'm thinking, I'd love for you to
share it, whether on this list, privately to me, to the 2015 Nomcom, or
to other Nomcom-eligible IETF participants when you're asking them to to
sign the recall petition ;-)

(*) http://en.wikipedia.org/wiki/Tug_of_war