On 1/15/15, 6:35 AM, "Bertrand Delacretaz" <bdelacre...@apache.org> wrote:
>On Thu, Jan 15, 2015 at 3:24 PM, Rich Bowen <rbo...@rcbowen.com> wrote: >> ...*historically*, when this kind of thing has happened (project >>implements, >> thing becomes critical), gradually it becomes the responsibility of >>Infra, >> not of the project, to do ongoing maintenance.... > >Yes, this is why I'm reluctant to encourage any initiative that >requires our infrastructure team to support new tools. And I suspect >infra shares that reluctance ;-) > >That being said, it's always a question of benefits vs. costs - but if >a simple thing using technologies that every web developer is supposed >to know works the choice is a no-brainer for me. IMO, that reluctance is the challenge faced by any ASF project that isn’t yet on the list of what everyone is “supposed to know”. AIUI, Infra is also staffed by volunteers so project folks can volunteer to be Infra for their project’s usage at the ASF. And if it isn’t “easy” for non-project Infra folks to grok how the project’s technology works, that is just another challenge for the project. One would suspect that they’ll face the same battle acquiring new customers anyway. I’d guess that most ASF projects are not yet a standard and want to be the new standard. Getting that first testimonial is often key to becoming the new standard. It seems wrong for the ASF to discourage establishment of pilot implementations until the project establishes a track record on some other set of customers. IMO, the ability to get an Azure VM is game changing in this regard. The project’s committers can take full responsibility for the pilot program. All Infra might need to do is establish one-way or two-way database or file mirrors so the pilot can’t mess up what works until it is deemed ready. I’d bet that most of a project’s customers would do that anyway. Infra should be encouraged to learn new things if the ROI is established by the pilot program and includes the cost of training non-project Infra folks, and those folks should ask for support like any other customer on that project’s list, and if they don’t get timely and helpful support, reject the product just like any other customer would. In summary, the ASF should be a slightly more willing customer for any of its projects. Azure VM’s seem to provide a way to do that without adding more load to Infra. -Alex
