Open ctb opened 9 years ago
I am a fan of the pull request hack.
This would make the use of submodules tempting to keep the docs next to the code. How about just split them prior to each release? That way the docs are still alongside the code during development.
Alternatively you could enable this on the main repo without splitting and use the new protected branches and required status checks to maintain sanity and have release be branches and not tags.
On Sat, Sep 05, 2015 at 01:37:20PM -0700, Michael R. Crusoe wrote:
I am a fan of the pull request hack.
This would make the use of submodules tempting to keep the docs next to the code. How about just split them prior to each release? That way the docs are still alongside the code during development.
-1 submodules; never heard anything good.
Keeping the docs up to date w/development is troublesome no matter how we do it; I don't see a clean solution either way (in separate tree, or in same tree).
Alternatively you could enable this on the main repo without splitting and use the new protected branches and required status checks to maintain sanity and have release be branches and not tags.
Sounds like management overhead, -0 from me.
To recap:
To enable the community to update the docs easier using the Pull Request Hack we are considering:
My response:
What is it we want to achieve with this?
On the idea of the PR hack: http://rfc.zeromq.org/spec:42/C4/ there is a lot of thinking in that document and I wonder if you need to adopt some of those ideas in addition to "give people commit rights".
From super small repos and bigger ones (like nteract/nteract
) it seems that giving people commit access mostly leads to them not using it. IMHO you need an existing community (or ability to quickly grown one) that agrees on what "high quality" is and "enforces" that, otherwise it seems to spiral out of control (-> some people on the project will feel like they spend all their time tidying up after others).
I'd be in favor of the PR hack, but would prefer that we kept the docs alongside the code.
With the move to Cython and oxli-ization (#1595) and people building on top of khmer (kevlar, and whatever @camillescott is doing) I'd like to revisit this. One compelling use case continues to be the ability to update docs independent of the need for a full khmer release.
Yeah, after a few months I'm starting to see the appeal of this.
heh
After 2.0 release, I'm thinking we should split the documentation off into its own repo and then try out the pull request hack:
http://felixge.de/2013/03/11/the-pull-request-hack.html
In that other repo, we could have branches that correspond to specific releases, with master corresponding to the latest master on khmer. So v2.0 would be the docs for khmer v2.0, but they could be updated routinely without going through a release.
Thoughts?