Open Stack

On Fri, Mar 25, 2016 at 03:20:30PM -0400, Doug Hellmann wrote:
> Excerpts from Jim Rollenhagen's message of 2016-03-25 10:45:30 -0700:
> > On Fri, Mar 25, 2016 at 09:10:05AM -0400, Doug Hellmann wrote:
> > > Excerpts from Lana Brindley's message of 2016-03-24 08:50:49 +1000:
> > > > On 24/03/16 08:01, Doug Hellmann wrote:
> > > > > Excerpts from Lana Brindley's message of 2016-03-24 07:14:35 +1000:
> > > > >> Hi Mike, and sorry I missed you on IRC to discuss this there. That said, I think it's great that you took this to the mailing list, especially seeing the conversation that has ensued.
> > > > >>
> > > > >> More inline ...
> > > > >>
> > > > >> On 24/03/16 01:06, Mike Perez wrote:
> > > > >>> Hey all,
> > > > >>>
> > > > >>> I've been talking to a variety of projects about lack of install guides. This
> > > > >>> came from me not having a great experience with trying out projects in the big
> > > > >>> tent.
> > > > >>>
> > > > >>> Projects like Manila have proposed install docs [1], but they were rejected
> > > > >>> by the install docs team because it's not in defcore. One of Manila's goals of
> > > > >>> getting these docs accepted is to apply for the operators tag
> > > > >>> ops:docs:install-guide [2] so that it helps their maturity level in the project
> > > > >>> navigator [3].
> > > > >>>
> > > > >>> Adrian Otto expressed to me having the same issue for Magnum. I think it's
> > > > >>> funny that a project that gets keynote time at the OpenStack conference can't
> > > > >>> be in the install docs personally.
> > > > >>>
> > > > >>> As seen from the Manila review [1], the install docs team is suggesting these
> > > > >>> to be put in their developer guide.
> > > > >>
> > > > >> As Steve pointed out, these now have solid plans to go in. That was because both projects opened a conversation with us and we worked with them over time to give them the docs they required.
> > > > >>
> > > > >>>
> > > > >>> I don't think this is a great idea. Mainly because they are for developers,
> > > > >>> operators aren't going to be looking in there for install information. Also the
> > > > >>> Developer doc page [4] even states "This page contains documentation for Python
> > > > >>> developers, who work on OpenStack itself".
> > > > >>
> > > > >> I agree, but it's a great place to start. In fact, I've just merged a change to the Docs Contributor Guide (on the back of a previous mailing list conversation) that explicitly states this:
> > > > >>
> > > > >> http://docs.openstack.org/contributor-guide/quickstart/new-projects.html
> > > > > 
> > > > > I think you're missing that most of us are disagreeing that it is
> > > > > a good place to start. It's fine to have the docs in a repository
> > > > > managed by the project team. It's not good at all to publish them
> > > > > under docs.o.o/developer because they are not for developers, and
> > > > > so it's confusing. This is why we ended up with a different place
> > > > > for release notes to be published, instead of just adding reno to
> > > > > the existing developer documentation build, for example.
> > > > > 
> > > > 
> > > > All docs need to be drafted somewhere. I don't care where that is, but make the suggestion of /developer because at least it's accessible there, and also because it's managed in the project's own repo. If you want to create a different place, or rename /developer to be more inclusive, I think that's a great idea.
> > > 
> > > I think we'll want to add a new location, or publish to a similar
> > > location to the existing install guide. I found, for example
> > > http://docs.openstack.org/mitaka/install-guide-ubuntu/
> > > 
> > > If we take ironic as the example, and assume all OS-types would be
> > > covered in one manual, we could make that:
> > > 
> > > (1) http://docs.openstack.org/mitaka/ironic/install-guide/
> > > (2) http://docs.openstack.org/ironic/mitaka/install-guide/
> > > (3) http://docs.openstack.org/install-guide/ironic/
> > > (4) http://docs.openstack.org/ironic/install-guide/
> > > 
> > > I like options 3 and 4. To keep things simple for the project teams,
> > > I think we want to skip including the release series in the base
> > > URL.  As the instructions change, projects may need to create
> > > separate sub-sections of their guide for each series, but they
> > > should be able to do that without having to set up separate publishing
> > > locations and jobs.
> > > 
> > > Another benefit of options 3 and 4 is that as the ironic team
> > > produces other guides, those can go under a consistent URL that
> > > makes it relatively simple to set up a generic publishing job for
> > > all projects. Option 4 does have the benefit of making it easy to
> > > have a page at http://docs.openstack.org/ironic/ linking to all of
> > > the guides the ironic team has published.
> > > 
> > > Thoughts?
> > 
> > I also like 3 and 4. I like 3 because it's a similar structure to the
> > developer docs, however I do like 4 giving us the option to publish
> > other guides (and perhaps we could move the developer docs to
> > /ironic/developer).
> > 
> > I do think we should be able to publish per-release versions of the
> > install guide (or any other guides) similar to how we publish developer
> > docs per-release today. For example:
> > http://docs.openstack.org/developer/ironic/liberty/
> > http://docs.openstack.org/developer/ironic/5.1.0/
> 
> Interesting, I wasn't aware of anyone doing that. Do you have custom doc
> build jobs in place? I could imagine this being useful for the Oslo
> libraries, too.

Nope, it just works. Random examples:
http://docs.openstack.org/developer/oslo.config/liberty/
http://docs.openstack.org/developer/oslo.config/mitaka/
http://docs.openstack.org/developer/oslo.config/3.4.0/

Just an undocumented feature of the docs job we found one day. :)

// jim

> 
> Doug
> 
> > This gives a canonical place for deployers to look for the guide for
> > their specific version, and solves the earlier piece about excluding the
> > release series from the base URL (which I agree with).
> > 
> > The downside being that intermediate releases (for ironic, today) cannot
> > be updated if there's something missing or wrong; however that can
> > easily be fixed by branching intermediate releases similar to what some
> > other projects (telemetry?) do, if we decide that's valuable (though I
> > don't like the idea of more backports to get a bug fix back to a given
> > series). This is already a known caveat of using these URLs today, so I
> > don't think it's worth blocking on. :)
> > 
> > My only questions are how much work this will be (including all the
> > redirects/rewrites that will need to happen), and where this happens (is
> > it all within the infra projects, or is there also docs work that will
> > need to happen?).
> > 
> > // jim
> > 
> > > 
> > > > > Right. The solution to that isn't to say "we aren't going to document
> > > > > it at all" or "publish the documentation somewhere less ideal",
> > > > > though, which is what it sounds like we're doing now.  It's to say
> > > > 
> > > > Actually, I said that I acknowledge that isn't working, and we need to find a different solution.
> > > 
> > > OK, that wasn't clear to me from your message that continued to suggest
> > > publishing to a location under /developer.
> > > 
> > > Doug
> > > 
> > > __________________________________________________________________________
> > > OpenStack Development Mailing List (not for usage questions)
> > > Unsubscribe: OpenStack-dev-request at lists.openstack.org?subject:unsubscribe
> > > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
> > 
> 
> __________________________________________________________________________
> OpenStack Development Mailing List (not for usage questions)
> Unsubscribe: OpenStack-dev-request at lists.openstack.org?subject:unsubscribe
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev