Open Stack

Le 2014-10-06 05:26, Tom Fifield a écrit :
> On 04/10/14 04:03, Nick Chase wrote:
>> 
>> On Fri, Oct 3, 2014 at 3:07 PM, Stefano Maffulli 
>> <stefano at openstack.org
>> <mailto:stefano at openstack.org>> wrote:
>>     >  1. Pick an existing topic or create a new topic. For new 
>> topics,
>>     we're
>>     >     primarily interested in deployment scenarios.
>>     >  2. Develop content (text and/or diagrams) in a format that
>>     supports at
>>     >     least basic markup (e.g., titles, paragraphs, lists, etc.).
>>     >  3. Provide a link to the content (e.g., gist on github.com
>>     <http://github.com>, wiki page,
>>     >     blog post, etc.) under the associated topic.
>> 
>>     Points 1-3 seem to be oriented at removing Launchpad from the 
>> equation.
>>     Is that all there is? I guess it makes sense to remove obstacles,
>>     although editing the wiki (since it requires a launchpad account 
>> anyway)
>>     may not be the best way to track progress and see assignments.
>> 
>> 
>> No, really, the main change is in step 5.  Launchpad isn't the 
>> problem,
>> as far as we can tell; Docbook is.
> 
> Hi Nick,
> 
> As best I can tell - 'step 5' has been in place for at least the last
> few summits at least, so this is not a change :) We have had a policy
> where anyone can dump text in bug reports and we'll wrangle it. This 
> has
> been popular, see eg Marco Cossoni's contributions, but in my opinion
> not widely enough communicated - so thanks for your efforts.

We actually have another way to work with developers, although it's 
been only available for the new HOT guide. This guide is temporary, it 
will become a part of the user guide. The interesting point is that it's 
written in RST, and uses gerrit for reviews. So far we've had 2 core 
members of the heat team contributing content, and this content has been 
reviewed by other members of the team.

The devs patches focused on content, not on the form of the content. I 
suggested to accept the patches rapidly - as long as they're technically 
correct - and to rework them later (what I've started to do a couple 
days ago). The fact that we're using gerrit and that the developers 
review each other work makes me more comfortable with the quality of the 
content.

I'd really like to see this process extended to a larger part of the 
documentation, although this might not be needed everywhere.

I had this workflow in mind:

* a dev sends a review to a temporary repo
* other devs can validate the information, and give their +1 when the 
patch is ready
* a doc reviewer either requests more technical detail, or gives his 
+2/accept
* the doc team reworks the patch and integrates it to the doc 
repository

I really think that the process worked for the HOT guide, and I'm 
convinced that it could work for other parts of the doc (Cinder and 
Neutron drivers doc for instance).

As a side note, we have a tool that converts RST to docbook. The hot 
guide is automatically built using this tool 
(http://docs.openstack.org/hot-guide/content/hot_guide_hot-guide.html).

Gauvain