My thoughts about the problems of the IETF

[Jonne.Soininen at nokia.com]

> > but isn't IETF last-call there for catching the inter 
> WG/inter Area issues? 
> 
> 1. it's far too late to be effective at catching such issues. 
>  you want to
> identify them early on, before a design is frozen

I think you are actually right in this. Maybe there should be some kind of "concept last-call" or presentation of the concept to a bigger audience early on. Some WGs have already started to do requirements documents that should contain the focus of the work. Maybe there should be one intermediate step of writing an architecture document, or concept document that kind of describes what is approach envisioned before writing the actual protocol. 

> 2. working groups cannot be trusted to be responsible at 
> evaluating issues
> from outside their narrow focus - or for that matter, to even 
> understand
> those issues and give them adequate consideration.

One thing what I would argue is that PS is as its name says "proposed standard". It could be seen as an intermediate phase towards a standard. This way the technical work has been a bit stabilized and the discussion can then focus on mending the actual protocol based on the PS. 

Anyway as I wrote to Brian as well - I sometimes tend to trust a bigger group of people to be more objective than a smaller group of people...

> 3. a really good way to fail to accomplish something vital is 
> to ask a few 
> thousand people to do it, and hope that somebody takes it up. 
>  everyone can
> then claim it's someone else's problem.
> 
> we need last call for the sake of openness and fairness, but 
> we shouldn't
> pretend that it's a good way to do reviews.  think of it as a 
> low-overhead
> appeal.
> 

I think there may also be a bit of a chicken and egg problem with the IETF last-call. As you know that IESG is going to take a look at it, and practically the IETF last-call is sometimes not that important anymore, people are not paying that much attention on them. If you know that this is really your last chance to comment, and nobody is doing it for you, you might take it more seriously.

> > Anyways, would you have some examples of such cases?
> 
> zeroconf is the most immediate example.  ipng is probably another.

I do not know zeroconf that well. However, I would be interested to know what you see as an example from ipng.  

> 
> > Maybe I'm too naive,
> > but I kind of tend to think that the WGs try to do the right thing. 
> 
> they usually try to do the right thing as they define it, but 
> that often
> is not the right thing for the community as a whole.  the 
> zeroconf people
> for instance were so focused on enabling local applications to operate
> without a configured network that many of them would ignore 
> repeated examples
> given to them of apps that this would break, claiming that 
> they didn't exist.

(Just for curiosity:) For instance?

Cheers,

Jonne.