[OpenStack-DefCore] Is DefCore like Linux Standard Base?

Monty Taylor mordred at inaugust.com
Mon Nov 2 10:45:38 UTC 2015


On 11/02/2015 05:04 AM, Thierry Carrez wrote:
> Meyer, Jim wrote:
>> [...]
>> The question here: how is DefCore different, and how will we succeed where LSB failed?

Great question. I think there is a lot to learn from the LSB.

> I think LSB failed at incentivizing complying distributions. Basicall
> you could still call your distro "Linux" even if you didn't follow the
> LSB. Defcore can (and does) use trademark programs to incentivize
> OpenStack distributions, so we should avoid that.

Yup. Unless the thing is pervassive, then consumers have to write the 
workaround code _anyway_ so the benefits are hard to achieve. For 
instance, lsb_release is a great program that is part of lsb, and we use 
it all the time, but check this out:

http://git.openstack.org/cgit/openstack-infra/system-config/tree/install_puppet.sh#n21

when we are bootstrapping a machine to get puppet installed so that we 
can do config management on it, we can't actually count on lsb_release 
being there because it's AMAZINGLY not part of the default minimal image 
on all of the platforms. We do install it first thing so that subsequent 
scripts can rely on it.

> Also the LSB was only one path to Linux interoperability. There are
> various ways to be completely interoperable without complying to the
> LSB. It is also no coincidence if LSB is beginning to die out at the
> same moment yet another way to interoperate at application bundling
> level appears (containers). I would argue that in the OpenStack case,
> Defcore is an essential piece of our interoperability story. There are
> others pieces (like killing differences in implementation when they
> bubble up to the user), but a common set of base services is a key element.
>

I think also that at least so far, our vendors value a standard to shoot 
for - and our tech community values enabling our vendors to adhere to 
it. Where there is a contentious issue that DefCore bubbles up, so far 
the response has been for everyone to get in a room and find an answer 
we actually can all get on board with. A good example of this is the 
Image Upload sessions we had at the summit based on DefCore not being 
able to find an image upload that all of our vendors and users were 
happy with. Guess what? I think we found the right compromise at the 
summit and will come out the other side with a thing that DefCore _can_ 
require because it's reasonable for all parties to implement.




More information about the Defcore-committee mailing list