Complex Problems

[James Kempf]

John,

> You raise a very important issue here.  If in 1996, there was a lot
> of input on architectural issues but last year there was very little -
> what is going on?  Is it that people don't care anymore, are the burnt
> out, or the stress of everyone's day job's is too much?  For some
> reason, it may be so that people are less committed to these principles.
> Do we need to re-energize folks on this or is it the communities feeling
> that this is someone else's problem - or could it just be that folks
> need to be educated on this?
>


I think the reason nobody gave any feedback is that the principles in RFC 1958
have been generally accepted and people don't see much need to change them,
with the exception of middleboxes, which Brian and others have dealt with in
other RFCs. Traditionally, in the past, the kinds of architecture done in the
IETF and specifically identified as "architecture" have been on the scale of the
entire Internet.

I believe many of the architectural problems faced by WGs today are not of the
large scale, Internet-wide sort that they have faced in the past, but are rather
micro-architectural. These typically masquerade as "frameworks" and have in the
past not been subject to dicussion as serious architecture, but rather are
examined on a piecemeal basis. For example, how to add a particular moderately
sized piece of functionality and have it properly integrated with the rest of
the Internet protocol suite. Another example is piecemeal solutions that have
been cobbled together by particular working groups for specific cases which,
when considered in a broader architectural context, actually intersect and would
result in a cleaner design if a common solution were developed. This kind of
complexity is fairly typical of a system as it matures. Defining the
interactions involved in adding new functionality, or in consolidating common
functions to simplify existing functionality, become more important than
articulating basic principles, because the basic principles have already been
articulated.

            jak