I like the idea to step back and I actually only now found the time to
read this thread, from its very start all the way to GDMO. (This is
how my email client mutt sorted the thread for me. ;-)

I agree that we do have a big big data model coordination challenge in
the IETF (and soon beyond as other SDOs start to extend IETF models).
The sad truth is that addressing the challenge requires say 85% human
brains (good early reviews paired with followup communication plus and
some oversight and steering) and perhaps 10% better tools and likely
only 5% conventions (if at all). This being the IETF, we seem to focus
on the 10% and the 5% of the challenge first (probably because we know
how hard it is to allocate the human resources needed to work on the
85% of the challenge). Things get somewhat interesting if well
intended attempts to address the 2% of the challenge end up
destabilizing data models that were written to form basic building
blocks and that we have published (through a consensus process) and
that have been implemented and deployed and that were also designed to
provide a basis for other SDOs to extend them (I am talking about
ietf-interfaces, obviously).

Concerning the openconfig proposal that started this debate, I fail to
see how good model design becomes significantly easier by having a
rigid fixed plan where the models are to be rooted. Writing good
extensible models is hard, and it will always be. A big part of the
problem is (i) to identify a common model for a technology that can be
implemented across multiple devices and (ii) to future prove the model
such that extensions of the model can work easily (and predicting the
future remains difficult for most of us).

To me, it seems we have a software engineering challenge paired with
the specifics of working in a volunteer organization that in addition
has its own internal boundaries (called areas and working groups). We
also witness the differences of opinion between people who believe
small design teams for criticial pieces can help to solve the 85%
challenge and people who believe more agile processes should be used
(where you throw out rough ideas quickly and subsequently iteratively
revise them to either become a good idea or to remove them if they did
not fly). Unfortunately, the meta-debate between a more traditional
approach and a more agile approach does not help with the 85%
challenge, at least not in the near future.

So in conclusion, I believe the key problem is a lack of skilled human
resources that can help with the 85% of the challenge. So what can we
do? Trying to increase the amount of skilled human resources involved
in data model writing (training, mentoring, sharing of know-how - but
this is a long-term process) and making sure that the skilled human
resources we have stay productive by focusing work, by prioritizing
work, and by maintaining a constructive atmosphere.

Looking at the many things going on and the number of people deeply
technically involved in those things, I am increasingly concerned that
we loose focus and make no measurable progress on the things we want
to achieve.

/js

-- 
Juergen Schoenwaelder           Jacobs University Bremen gGmbH
Phone: +49 421 200 3587         Campus Ring 1 | 28759 Bremen | Germany
Fax:   +49 421 200 3103         <http://www.jacobs-university.de/>

_______________________________________________
netmod mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/netmod

Reply via email to