Date: Thu, 30 May 2002 23:13:24 -0500
From: Dave Crocker <dhc2@dcrocker.net>
Message-ID: <5.1.0.14.2.20020530230656.02c2f690@jay.songbird.com>
| So, what exactly do folks think is a practical kind of change to the
| current IETF policies?
Actually, like many things, I suspect that the underlying problem
that we're seeing is in quite a different area than the one we're
looking in.
My suggestion to fix this problem is quite simple
No more last calls before moving protocols to historic,
except where they're full standards already.
For everything else, going to historic should be automatic. That is,
the only way to avoid any spec being made historic automatically, is
for it to become a full standard.
This is more or less saying what Henning Schulzrinne said before I got
the chance...
The problem is that very few standards make it to Draft.
Exactly. We have developed a culture of getting the work to PS
status, and considering it done. We even disband working groups
(or move them to dormant status, which is effectively the same thing)
as soon as all their docs have been published.
And the implementors see that - as soon as something has reached PS
status, it is considered finished, and we even have people getting
upset (because of the deployed base we'll be breaking) if any changes
get made.
That's what's really broken here - we actually have almost the prefect
IPR policy in place, but we're not bothering to actually attempt to
use it before it is way too late.
What we need to do is make sure that everything gets to draft standard
within a relatively short time after it has reached PS, say 9-12
months (preserving the current 6 month minimum to make sure there's
enough time for implementations to be attempted). Anything that
hasn't made it to DS then should simply be shelved, abandoned (DS isn't
much of a hurdle after all). And we should be actively discouraging
distribution in products of anything that isn't at least DS (to do
that we probably should get into the habbit of making random innocuous
changes to docs when they go to DS state - like if we have commands
that are issues in binary and they're 1, 2, 3, ... (as usual), we
just change which is 1, which is 2, etc so everyone actually running
the code has to update - which of course is impossible after something
has been shipped.
Then again, we need to make sure that all DS's get elevated to full
standard quickly, or dropped. This is where the IPR rules really
get exercised, where we see if the proposal has IPR rules that make
it difficult to implement and deploy or not. If the IPR rules cause
problems, then the doc won't reach Std status, and again, that should
mean that it is automatically made historic and abandoned.
It isn't the IPR rules that need changing here, the current formulation
that Christian Huitema came up with is almost certainly the best we
can do - anything different is far more likely to cause problems than
to solve any.
kre