Open Stack

On 10/30/2012 12:38 PM, Eric Windisch wrote:
>>
>> 2) It implies that this is a stable, supported API.
>>
>> This one is probably more controversial. Configuration options are
>> knobs exposed to users of OpenStack. An option that takes a class path
>> says "Hey, feel free to plug in whatever code you want right here!" I
>> don't think that's what we really want. If it *is* what we want (I hope
>> not), we've got a whole new set of issues to discuss. If we have a
>> public API, we need to do it right and talk about the rules for
>> maintaining stable APIs for these plug points.

> Cloudscaling runs stock OpenStack without forking the code. This is
> accomplished through loading modules passed in through the various
> backend_driver options. It is essential that we can plug into these
> places to avoid forking.
> 
> I think entrypoints should be satisfactory, but the devil does seem to
> be in the details. I'll have to dig deeper into Monty's patches.

I would argue that you *are* forking the code, even if not doing so
without modifying the code.

I also don't see what's so hard about carrying patches.  Packagers have
mastered this.  If you have custom code, I presume you have custom
packages, too.

-- 
Russell Bryant