> On Jan 10, 2016, at 12:50 PM, Daan Hoogland <daan.hoogl...@gmail.com> wrote: > > people, what we are saying by moving to github, completely is that, for > now, we will be moving away from apache. Apache can not take responsibility > for any code that is not under it's control.
Daan, thanks for jumping right in. Part 1: -------- As VP I started this thread and clearly I am *not* advocating leaving Apache. - They are other Apache projects asking to host their repo on github with full access to github features. We are not alone. - There is an experiment going on with the Apache Whimsy project to see how being hosted on GitHub could be accomplished and still fulfill requirements of the ASF. The board is aware of these demands and is trying to get data to see how this could be done. - What I am looking for in this discussion, is to get a consensus in our community on whether we would like to make this move or not. Leave aside the problems. Imagine an ASF project can still be an ASF project even if hosted on GitHub. Would we do this ? Part 2 : ——— There are indeed issues with such a move, but I think the lines are more blurry than we think. 1-ASF projects already used proprietary software (hosted or not hosted on ASF infra, think JIRA, Hipchat, Slack) 2- Moving to GitHub would require all committers to get accounts on GitHub, so we would be forcing folks to get registered with a third party hosted service. That said we already do this, since this is our commit workflow. 3- ASF would loose control on push logs (this is the main pain point for the board, as far as I can tell). This means that if there was any problems/inquiry from lawyers about who made which commits (provenance of the code), ASF may not have the info required. 4- The line is blurry, until couple years ago, ASF did not keep all push logs -even for svn-, so inquiries mentioned in 3- would not be able to be answered anyway (this is my understanding, this may not be 100% accurate) 5- ASF seems to keep some logs on IP addresses of commits, a potential privacy law issue, especially in the EU. 6- Git commits can be PGP signed, so committers could commit with their ASF PGP keys, I don’t see what better source of provenance you could find. Part 3: ———— To me the main issue for us is that our current privileges on GitHub prevent us from building more productive CI workflow and makes the life of the RM more difficult (cannot use labels, cannot use issues, cannot configured triggers/hooks etc). We want things to be easy and should work on removing barriers to productivity. That’s the reason I am advocating that we make such a move. Indeed > This is fine, but to satisfy > the foundation policy, maintaining an *Apache* CloudStack we would have to > have a process in place for contributing back to the foundation. We have > been moving at an unmaintainable pace the last few month so change has to > happen. > > Please let's discuss this carefully and with tempered down emotions. I am > about +e -π for going to github and would just not know what move to github > entails off the top of my head. One nice challenge is see is creating a 32G > docker containing a cs and two hosts that can be spun up and run a set of > integrations tests on a PR-trigger. Perfectly doable. Any volunteers? It > can be started outside of the move to github because it can be triggere as > we speak from any ci implementation. > > > But this is not the only one. Let's define as much as possible small > improvements and implement them. These will help us whether we move to > github or not. > > regards, > Daan > > On Sat, Jan 9, 2016 at 3:25 AM, chunfeng tian <tianyunqu...@gmail.com> > wrote: > >> +1 to moving to github. >> >> On Wed, Jan 6, 2016 at 6:17 PM, ilya <ilya.mailing.li...@gmail.com> wrote: >> >>> +1 to moving to github. >>> >>> On 1/3/16 3:25 AM, Sebastien Goasguen wrote: >>>> Bringing this one discuss thread to the top of the ML to get stronger >>> consensus. >>>> >>>> We need it if we want to request a move to GitHub. >>>> >>>> Note that this is not about leaving the ASF, it is about using GitHub >> to >>> its full potential. >>>> >>>> The ASF board is investigating ways for a project to use Github and >>> still maintain strong provenance of commits to keep the high quality and >>> provenance standards of ASF code. >>>> >>>> If we get consensus we can request to the board to be part of the >>> “trial” and move to Github. >>>> >>>>> On Dec 21, 2015, at 11:37 AM, Sebastien Goasguen <run...@gmail.com> >>> wrote: >>>>> >>>>> >>>>>> On Dec 21, 2015, at 11:34 AM, Daan Hoogland <daan.hoogl...@gmail.com >>> >>> wrote: >>>>>> >>>>>> Sebastien, This will create a github repo under the apache >> organisation >>>>>> right? one that we can not merge to. >>>>>> >>>>> >>>>> Yes , that’s how I created all the docs repo and the repos for >> ec2stack >>> and gstack. >>>>> >>>>> >>>>> >>>>>> On Mon, Dec 21, 2015 at 10:51 AM, Sebastien Goasguen < >> run...@gmail.com >>>> >>>>>> wrote: >>>>>> >>>>>>> BTW >>>>>>> >>>>>>> Anyone can ask for a new git repo which will be mirrored on github >> at: >>>>>>> >>>>>>> >> https://issues.apache.org/jira/servicedesk/customer/portal/1/create/8 >>>>>>> >>>>>>> Not sure if the link will work, but it’s available through issues. >>>>>>> >>>>>>>> On Dec 19, 2015, at 7:03 PM, Sebastien Goasguen <run...@gmail.com> >>>>>>> wrote: >>>>>>>> >>>>>>>> >>>>>>>> >>>>>>>>> On 19 Dec 2015, at 16:28, Rene Moser <m...@renemoser.net> wrote: >>>>>>>>> >>>>>>>>> Hi Seb >>>>>>>>> >>>>>>>>>> On 12/19/2015 10:12 AM, sebgoa wrote: >>>>>>>>>> >>>>>>>>>> Late October I started thread [1] about moving our repo to >> GitHub, >>> I >>>>>>> would like to re-open this discussion. >>>>>>>>>> >>>>>>>>>> Now that we have stabilized master and release 4.6.0, 4.6.1, >> 4.6.2 >>> and >>>>>>> 4.7.0 we need to think about the next steps. >>>>>>>>>> >>>>>>>>>> To me Git and GitHub has become an essential tool to any software >>>>>>> development, not using it to its full potential is hurting us. >>>>>>>>>> >>>>>>>>>> Just as an example I would like to point you to [2], this a PR I >>> made >>>>>>> to Kubernetes (a container orchestrator), it literally added 14 >>> characters >>>>>>> in a json file. >>>>>>>>>> This was really a very minor change. >>>>>>>>>> >>>>>>>>>> The PR automatically triggered 3 bots which created 7 labels, it >>> ran >>>>>>> end to end testss, Jenkins jobs and triggered third part builds. >>>>>>>>>> It was automatically merged. >>>>>>>>> >>>>>>>>> I am fine moving to github. >>>>>>>>> >>>>>>>>> But IMHO the git hosting is not the problem, the problem is how >> far >>> do >>>>>>>>> we trust the current tests and how we can them improve. >>>>>>>>> >>>>>>>>> Moving to github doesn't improve testing. Doing manual tests is >>> okay and >>>>>>>>> hard work, it does not speed up things. >>>>>>>>> >>>>>>>>> We need fully automated unit _and_ integration tests that we >> trust. >>> I do >>>>>>>>> not trust in mocking and simulating infrastructure. >>>>>>>>> >>>>>>>>> We discovered most of the major problems running cloudstack on >> real >>>>>>>>> hardware in real world scenarios. Race conditions, unexpected VR >>>>>>>>> reboots, VMs not getting IPs from DHCP, etc. >>>>>>>>> >>>>>>>>> Rating complexity of changes: easy_fix, minor_change, major_change >>>>>>>>> >>>>>>>>> Running tests according complexity: >>>>>>>>> >>>>>>>>> - easy_fix: just merge it. >>>>>>>>> - minor_change: unit and simulator test passed >>>>>>>>> - major_change: the full blown integration testing >>>>>>>>> >>>>>>>>> IMHO we should work on solid testing and development is fun, >>> merging a >>>>>>>>> click and releasing a breath. >>>>>>>>> >>>>>>>>> Just my 2 cents. >>>>>>>> >>>>>>>> Fully agree >>>>>>>> >>>>>>>> I do think moving to github would allow us to run tests on real >>> systems >>>>>>> more easily. >>>>>>>> >>>>>>>> >>>>>>>>> >>>>>>>>> Regards >>>>>>>>> René >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>> >>>>>>> >>>>>> >>>>>> >>>>>> -- >>>>>> Daan >>>>> >>>> >>> >> >> >> >> -- >> Tian ChunFeng >> http://cloud.domolo.com >> > > > > -- > Daan
