Narayan Desai wrote:
I think there are two ways that discussion of current tools can be
interesting. I don't think that features are necessarily so
interesting, but the design decisions and goal can be fascinating. For
example, many of bcfg2's compelling features are rooted in design
decision to base everything around a validation model. While the
features are compelling, the real contribution it has to offer is that
if you build a system around validation, you have these additional
options available with which you can build interesting
functionality. I would say on the research side, the why is as, if not
more, important than the what.
I agree, and I would go one further. I'm sure there are some problems
that Bcfg2 is better at than Puppet, and vice-versa. Being able to talk
about design decisions and their uncumbent trade-offs with high-level
practitioners gives them the opportunity to understand what went into
the tools and how to decide between them, and it gives us tool
developers more information for either making the next trade-off or
rethinking previous decisions.
I'd like to know more about what you mean when you say "designed around
validation". I think that Puppet and Bcfg2 have similar features in
terms of granular reporting of how configurations are out of sync and
what needs to be changed, but we've clearly gone different directions in
terms of how to take advantage of this. I'd like to spend some time
studying why.
Secondly, I think that many configuration management papers are
missing practical grounding. (ie, they have the "I built this thing",
but are missing the "the users hated it for what we intended, but
loved it for something completely different") For example, giving the
bcfg2 user base the above features has proven quite interesting, as
they've starting using the tools in ways I never would have
predicted. I find these cases to be more telling than the cases where
we implemented a solution to a problem and the users used it precisely
as we had intended.
Heh, I think that was Larry Wall's definition of success -- when people
start doing things with your product that you never intended. I agree.
--
I take my children everywhere, but they always find their way back home.
--Robert Orben
---------------------------------------------------------------------
Luke Kanies | http://reductivelabs.com | http://madstop.com
_______________________________________________
lssconf-discuss mailing list
lssconf-discuss@inf.ed.ac.uk
http://lists.inf.ed.ac.uk/mailman/listinfo/lssconf-discuss
