DevOps - SVN Workflow

Flickr
I've read about DevOps  a wikipedia and one aspect of it is not clear for me. How in Flickr they use Subversion? At this presentation: http://www.slideshare.net/jallspaw/10-deploys-per-day-dev-and-ops-cooperation-at-flickr
page: 38 - Always ship trunk
page: 40 - Example of "branching in code"
And there is Trunk for development, release branch to support etc. But how it looks like in Flickr? There is at page 38 "Always ship trunk" so developers must work on branches most of the time? Or first commit is with 'branching if's" in code?
Personally I doesn't like a if's for branching idea...

Chrome
Recently I read about similar approach in Google Chrome Release Cycle in which workflow are embedded   configuration options to do branching in code.
1: Everyone works on trunk, then they create stable branch and work on dev and to stabilize branch and ship all scheduled functions.
2: They work on trunk, has 3 channels, trunk, beta and stable to work and are using branching in code to minimize merges.
But Chrome is a big project and it is still app. For now I think that for smaller projects 3 lines are to much, and I still think that I prefer SVN branches that branches in code...

Standard
At least I found it quite popular and used in couple of books as a 'good practice'.
Despite Chrome, standard SVN repository structure is like : http://ariejan.net/2006/11/24/svn-how-to-structure-your-repository/.
This is great, but existing site development is much more like at Flickr, small changes but often to make users feels that site from day to day has new features or at least team is fixing bugs fast :-]

Combine them

Now I think about approach similar to both of them( Chrome(1) flow + Flickr + Standard).
1: Everyone commits to trunk
2: There is stable branch that changes are merged to when ready
3: Big changes creates new stable branch(release 2.0)

And when I think about changes:
1-2 is about small tweaks and bugfixes like in Flickr
3 is about bigger changes, eg change of database to NoSQL, new website design or at least something to have opportunity to clean-up stable branch and merge it with trunk :)

Maybe I am reinventing the wheel, please comment and let me now what you think about it :)

Comments

Popular posts from this blog

How to simulate slow connection (developer proxy in Node.js)

How to use NPM packages from private repositories on bitbucket