On 11/15/2015 11:41 PM, Wouter Verhelst wrote: > On Sun, Nov 15, 2015 at 11:28:57PM +0100, Thomas Goirand wrote: >> On 11/15/2015 06:50 PM, Wouter Verhelst wrote: >>> On Thu, Nov 12, 2015 at 09:04:29PM +0100, Thomas Goirand wrote: >>>> On 11/12/2015 04:52 PM, kuLa wrote: >>>>> On 2015-11-12 15:58:03, Thomas Goirand wrote: >>>>>> As per the discussions during debconf, to be called "official", the >>>>>> images have to be built: >>>>>> - directly from an unmodified stable >>>>>> - with reproducibility on any Debian computer (ie: no need for any >>>>>> external infrastructure access) >>>>> >>>>> I don't think we reached any consensus in relation to the last point but >>>>> I'm >>>>> not going to argue about it right now. >>>> >>>> There's IMO no consensus to have, unless we change the root of Debian >>>> (ie: the DFSG, and the fact that we do free software, and can build it >>>> in Debian). The need for an external infrastructure would make the >>>> images non-free. SaaS on a proprietary platform is as non-free as one >>>> can get. I don't anyone would say otherwise, would you? >>> >>> Personally, I disagree with the statement that "the need for external >>> infrastructure would make the images non-free". >>> >>> If a cloud platform does not make it possible to *import* images from an >>> external source, >> >> AFAIK, they all do. >> >> If you can't upload a custom images, then it makes the cloud pretty much >> useless. > > Regardless, I still think you should state the requirement in terms of > the result, not in terms of how it's built. The "import image" thing was > just an (arguably bad) example; but I can imagine other cases where > things "have" to be built on the cloud provider's infrastructure, in the > sense that for practical reasons it's the best way forward, but that it > isn't theoretically the only possible way of doing so. > > In that light, I think a requirement that it be *possible* to build a > bit-for-bit identical image while outside the given cloud infrastructure > (even if for whatever practical reason we end up not doing so) makes > much more sense.
Then we do agree. That's what I wanted to write. Thomas Goirand (zigo)
