Theodore Ts'o tytso at
Wed Feb 26 10:22:48 CET 2003

On Tue, Feb 25, 2003 at 09:42:52AM -0800, Dave Crocker wrote:
> That is, a PS is allowed to tackle a very narrow problem, if the PS will
> in fact do something useful.  Narrow scope makes it easier to do things
> in a more timely fashion and with a better understanding of what is
> being done.

In general, I would agree with this, with some caveats. First of all,
we need to differentiate between "narrow scope" and "non-extensible".
Some of the better success stories as far as IETF protocols are
concerned are ones where it was relatively easy to add extensions to
the base protocol afterwards: PPP, DNS, and so on.

That being said, I will certainly admit that disasters can result by
making protocols too extensible.  And I certainly don't want anyone to
use the above paragraph as a justification for using ASN.1 ("friends
don't let friends use ASN.1") because of it's alleged extensibility.

Secondly, it MUST be the case that certain key features, like
security, have to be designed in from the start, and not added in as
an afterthought later.  Otherwise, you end up with design mistakes
such as presupposing a gobal access control table in every single core
router, specifying which x.509 certificate is authorized to speak for
a particular IP address.

						- Ted

More information about the Problem-statement mailing list