You all rock and our core/owners[1] list has almost doubled since our last discussion[2] about finer-grained owners files in April 2013! In light of how core/owners is being used I would like to revisit the discussion with the following changes:* Add owners files for the subdirectories that have turned out to be too coupled with core to become modules but still have implicit owners. These sub-owners files would follow the standard chromium owners policy outlined in http://www.chromium.org/developers/owners-files.
* Add comments to the current owners file where sub-owners files are not appropriate. This will clean up some unnecessary friction when finding a good reviewer. Owners should TBR changes to add comments above their name.
I've found a shortage of owners in web/ to be a bottleneck. Any chance we could revisit our policy for ownership there, too? (Would it be crazy to look at public/* ownership, too?)
To unsubscribe from this group and stop receiving emails from it, send an email to blink-dev+...@chromium.org.
Seems like a good case for comments, so that it's easier to tell which reviewers are best for DOM, CSS, etc. Even if that list remains in core/OWNERS.
As a committer outside of PST, it would be nice if we could keep the timezone distribution of OWNERS in the new scheme. It is always nice to have a local OWNER for better throughput.
How do people feel about sub-directory owners, one year later?I've found the places that have sub-directory OWNERS useful (I've pointed people to core/html/parser/OWNERS when they were looking for parser experts, looked to core/animation/OWNERS to remember who owns animation, core/editing/OWNERS for editing, etc.), and wish it existed in other places that were reasonable (for instance, I have a mental list of CSS experts, but an explicit core/css/OWNERS would be nice).Even if they are just subsets of core/OWNERS and platform/OWNERS (i.e. they are purely informational), I'd like to suggest that other directories in Blink adopt a similar policy. (In particular, I'm interested in creating core/paint/OWNERS and platform/graphics/OWNERS.)
In Chrome we do this quite often, in part because we don't really have a concept of many people having core ownership, but also in part because often people's domain knowledge is specialized and it's appropriate to have someone in a narrow directory but not the wider root.I actually am not a good core owner in Blink. I'm listed in Source/core/OWNERS, but I really shouldn't do code reviews for non-image decoding stuff.
When it comes to that stuff, in certain areas I'm the expert, so it wouldn't be appropriate to remove me from OWNERS entirely. It would be best to simply list me in the right subdirectories and not a root file.I think Blink's current idea of having many people in core OWNERS files is a mistake, because it doesn't really represent how people's knowledge and abilities actually map to the code organization.
On Wed, Sep 16, 2015 at 3:08 PM, Peter Kasting <pkas...@google.com> wrote:...In Chrome we do this quite often, in part because we don't really have a concept of many people having core ownership, but also in part because often people's domain knowledge is specialized and it's appropriate to have someone in a narrow directory but not the wider root.I actually am not a good core owner in Blink. I'm listed in Source/core/OWNERS, but I really shouldn't do code reviews for non-image decoding stuff.You're welcome to remove yourself from core/OWNERS and just ask for rubber stamps instead.
When it comes to that stuff, in certain areas I'm the expert, so it wouldn't be appropriate to remove me from OWNERS entirely. It would be best to simply list me in the right subdirectories and not a root file.I think Blink's current idea of having many people in core OWNERS files is a mistake, because it doesn't really represent how people's knowledge and abilities actually map to the code organization.Our system requires a higher bar, and necessarily so because the web platform API is very vast and the complexity of the engine is large.
On Wed, Sep 16, 2015 at 3:18 PM, Elliott Sprehn <esp...@chromium.org> wrote:
When it comes to that stuff, in certain areas I'm the expert, so it wouldn't be appropriate to remove me from OWNERS entirely. It would be best to simply list me in the right subdirectories and not a root file.I think Blink's current idea of having many people in core OWNERS files is a mistake, because it doesn't really represent how people's knowledge and abilities actually map to the code organization.Our system requires a higher bar, and necessarily so because the web platform API is very vast and the complexity of the engine is large.
I'm not sure I understand the counterargument. I'm suggesting having fewer core OWNERS, not more. The bar to be in some root-level OWNERS file should be higher than it is, not lower. Coupled with that, I wouldn't gate that decision on putting someone in a leaf directory OWNERS file on whether they would be appropriate for one of the core files. The result of these two is that more people should be in more non-core OWNERS files, fewer in core files, and engineers looking for reviews should generally not look in core files unless implementing changes that couch large cross-sections of the codebase.
If you're not trusted enough to be in core/OWNERS, you're not trusted enough to be in core/layout/OWNERS since you have access to all the same API surface there. Just adding people who work on a silo has not worked historically, it just means that slio'ed team dumps lots of code into their silo touching the rest of the system since it makes reviews easier.To address this we're adding API boundaries, better ASSERTs, and include guards. Then we can discuss adding fine grained owners. :)