Definition of power and responsibility [Re: Delegationofpower(wasRE: Section 2.4 ofdraft-ietf-problem-statement-00.txt)]

[Eric Rescorla]

Brian E Carpenter <brian at hursley.ibm.com> writes:
> Eric Rescorla wrote:
> > I hear that a lot, but I'm not sure why people believe that,
> > except for the usual argument about the devil you know. Moreover,
> > I'm starting to think that the "no formal existence" thing is part
> > of the problem as well.
> 
> Can you be specific?
Well, it makes it very hard for the IETF to actually negotiate
business relationships, for starters. It also encourages us
to think of people with what are very close to full time management
jobs (the IETF chair and to some extent the ADs) as volunteers.

> Because changing this (e.g. by making IETF
> a legal component of ISOC, or by separately incorporating IETF) 
> would be a step with tremendous indirect consequences.
What do you think those consequences would be?

> > > Correct. Now the largest area today has 27 WGs and two ADs. I agree
> > > it seems like too many, even for management by exception. But I'm quite
> > > resistant to concluding that junior ADs would fix this problem, since
> > > this is not a military command structure.
> > I'm not sure what it not being a military command structure has
> > to do with it. The principles of hierarchy and staff management
> > are pretty standard in business as well.
> 
> Well, most companies are flattening their hierarchy these days,
> and at least in my company, matrix relationships are the rule
> rather than the exception. (A.k.a. getting rid of silos.)
There's a debate in the organizational management world about how
large the maximum span of control is. As a consequence, many
organizations who had adopted hierarchies based on small spans of
control are flattening. But the debate spans about an order of
magnitude (2 to 25). Noone thinks it's 100, which is the span of
control of an IESG member.  (This depends on believing that the WG
chairs aren't first-level managers, but my experience with WGs is
extremely consistent with that).

> In fact, many of the problems that come up too late in our
> documents seem to be ones that result from *missing* matrix
> relationships between WGs. Adding hierachy increases the risk.
Not necessarily. Hierarchy isn't introduced into systems for no
reason. It's introduced because the current system is overloading the
managers. If the managers are overloaded, they can't handle the matrix
relationships, even if they're in principle situated to (provided they
had enough time). The entire point of introducing hierarchy is to
increase the effective surface area of the managers. Obviously,
there's a tradeoff between tight coordination and manager overload,
but I think it's pretty likely that at the moment overload is the more
pressing problem

-Ekr

-- 
[Eric Rescorla                                   ekr at rtfm.com]
                http://www.rtfm.com/