[dpdk-moving] Proposal a Committer model

[Stephen Hemminger]

On Fri, 18 Nov 2016 10:45:39 +0000
Hemant Agrawal <hemant.agrawal at nxp.com> wrote:

> > Sent: Thursday, November 17, 2016 7:36 PM  
> > > On Nov 17, 2016, at 7:43 AM, Thomas Monjalon  
> > <thomas.monjalon at 6wind.com> wrote:  
> > >
> > > 2016-11-17 09:27, Mcnamara, John:  
> > >> From: Jerin Jacob [mailto:jerin.jacob at caviumnetworks.com]  
> > >>> I believe the multi-committers model may not fix current consensus
> > >>> slowness issue. Instead, if we are focusing on reducing the workload
> > >>> of Thomas, then I think git pull request based scheme will reduce
> > >>> the workload.  
> > >>
> > >> So, something like a Gerrit model?  
> > >
> > > No, the mechanics of committing is not time consuming.  
> > 
> > The time consuming part is the reviewing the patch and using Gerrit does
> > attempt to make sure the reviewers are emailed directly. This to me helps to
> > require reviews as sometimes the huge email volume on the list is difficult to see
> > patches someone needs to review.
> > 
> > I would much more prefer one email (from Gerrit) per patch set instead of 10 or
> > 30 emails in some cases for a single email. Using gerrit also combines the patch
> > review comments, patchwork and email list into one tool plus it can kick off the
> > build and checkpatch processes.
> > 
> > Having a Gerrit model is not a bad process model and allows for multiple
> > committers. I know it is a new tool and it is pretty simple to use.  
> 
> [Hemant] I also agree that Gerrit is much better from committer, reviewer and submitter prospective. 
> 
> Why are we having reservations in moving to Gerrit? 

Because Gerrit has many flaws:

https://kernel-recipes.org/en/2016/talks/patches-carved-into-stone-tablets/