[openstack-dev] [all] Embracing new languages in OpenStack
flavio at redhat.com
Thu Nov 10 17:23:58 UTC 2016
On 10/11/16 15:54 +0000, Chris Dent wrote:
>On Thu, 10 Nov 2016, Clint Byrum wrote:
>>Excerpts from Chris Dent's message of 2016-11-09 11:14:32 +0000:
>>>Let's just get on with making stuff and work out the problems (and of
>>>course there will be many, there always are) as they happen. That's
>>>what we do.
>>Apologies for the stretched analogy. What you're suggesting is that we go
>>build without building codes and without a city plan, because there's a
>>market force that we must capture. But when we let the tenants move in
>>(the operators) and the other tenants sewers back up and their roads
>>are backed up with traffic, we'll just deal with that later. I don't
>>think that's fair to anyone.
>Sorry, I think you are extrapolating _far_ too much from what I've
>said. I've not said we should have no plan, no process, anarchy,
>cats and dogs living together. I've said we should work out the
>problems as we go.
>The current process to create a plan (nevermind actually following
>it) is weighted (to me) too far in the direction of trying to be
>prepared for as much as possible. That's a recipe for getting it
>wrong; because we are humans who can't see the future.
>>I don't think we have to have a PERFECT plan, but we need to _acknowledge_
>>that this is the price of diversity and expansion. I personally think
>>it's worth it, but let's own the chaos and at least start with a rough
>>hypothesis of how each new language might effect the overall system, and
>>a plan to measure and react quickly.
>Yes, let's talk about it and acknowledge it. There will be a price
>of diversity and expansaion, yes. I'm very happy to hear that you
>agree it is worth it. I'm also happy to hear you acknowledge that
>it will be chaotic.
>However -- to push the current thread further into the extreme than it
>actually is, to make a point -- we don't need to reimplement oslo in
>every candidate language before we actually build and distribute
>something useful with real users in whatever that language might be.
>We've got to have some faith in ourselves as developers, users and
>operators that we have the capacity to adapt and learn for sake of new
>stuff that is good.
If you mean Oslo as the team that manages the shared code across projects then
yes, I think we have to and we need a way for sharing code. If instead you mean
oslo libraries then no, I don't think we need to re-implement them all in the
new language. What we need, however, is to guarantee compatibility for operators
and other devs. Having inconsistencies in the way config files are managed, the
RPC protocol is implemented, messages are logged is a recipe for a chaotic
adoption of the new language. This is what I'd like us to avoid.
There's some work that needs to be done up-front. I agree the discussion of
adding new languages have nnot been perfect but that's also what we do. We
evaluate our options, we study, we work out a way to embrace change and
innovate. That's what I'm trying to do here, work out a way for us to embrace
new languages and move on to the next change.
As Clint put ir, we need to own the change and the chaos that will come with it.
Not having these processes would just make it unmanageable for everyone.
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 829 bytes
Desc: not available
More information about the OpenStack-dev