Thierry Carrez wrote:
Joshua Harlow wrote:
[...]
* Be opinionated; let's actually pick *specific* technologies based on
well thought out decisions about what we want out of those technologies
and integrate them deeply (and if we make a bad decision, that's ok, we
are all grown ups and we'll deal with it). IMHO it hasn't turned out
well trying to have drivers for everything and everyone so let's umm
stop doing that.
About "being all grown-ups and dealing with it", the problem is that
it's mostly an externality: the choice is done by developers and the
cost of handling the bad decision is carried by operators. Externalities
make for bad decisions.
Fair point, so how do we make it not a externality (I guess this is
where the core services arch-wg thread comes in?)? It all reminds me of
the gimp and how its UI was really bad to use for years. Sometimes
developers don't make the best decisions (and rightly so I will admit I
sometimes don't either).
I agree that having drivers for everything is nonsense. The model we
have started to promote (around base services) is an expand/contract
model: start by expanding support to a couple viable options, and then
once operators / the market decides on one winner, contract to only
supporting that winner, and start using the specific features of that
technology.
Is it possible to avoid the expand/contract? I get that idea, but it
seems awfully slow and drawn out... I'd almost rather pick a good enough
solution, and devote a lot of resources into making it the best solution
instead of choosing 2 solutions (neither very good) and then later
picking 1 (by the time that happens, someone that picked solution #1
would be quite a bit farther ahead of you).
The benefit is that the final choice ends up being made by the
operators. Yes, it means that at the start you will have to do with the
lowest common denominator. But frankly at this stage it would be awesome
to just have the LCD of DLMs, rather than continue disagreeing on
Zookeeper vs. etcd and not even having that lowest common denominator.
On a side note, whenever I heard operators or developers it makes me
sad... Why do we continue to think there are two groups here? I'd almost
like there to be some kind of rotation among *all* openstack folks where
say individuals in the community rotate between companies to get a feel
for what it means to operate & develop this beast.
Perhaps some kind of internship like thing (except call it something
else), I'd certainly like to break down these walls that continue to be
mentioned when I don't really think they need to exist...
* Leads others; we are one of the older cloud foundations (I think?) so
we should be leading others such as the CNCF and such, so we must be
heavily outreaching to these others and helping them learn from our
mistakes
We can always do more, but this is already happening. I was asked for
and provided early advice to the CNCF while they were setting up their
technical governance structure. Other foundations reached out to us to
discuss and adopt our vulnerability management models. There are a lot
more examples.
Is it theortically possible that we just merge with some of these
foundations? Aren't we better as a bundle of twigs instead of our own
stick?
"A single twig breaks, but the bundle of twigs is strong." - Tecumseh
Why aren't we leading the formation of that bundle?
[...]
* Full control of infrastructure (mostly discard it); I don't think we
necessarily need to have full control of infrastructure anymore. I'd
rather target something that builds on the layers of others at this
point and offers value there.
+1
__________________________________________________________________________
OpenStack Development Mailing List (not for usage questions)
Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe
http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
