Hello all, I have concerns regarding the work the cloud team has been publishing wearing the label of Debian/Wheezy which at this time of discussion is currently stable. I have voiced these concerns on topical threads in this forum to stop what I considered to be a change that is in no way acceptable and consensus has agreed that the change should not be made (specifically that automatic updates should be enabled in the Debian/Wheezy labeled GCE image). That issue is not the root of the problem and having consensus that automatic updates should not be happening isn't actually the correct way to handle this situation. The fact that consensus of the cloud team prevented the change is the problem because changes like that should not even be possible to stable. I've been trying to make my point and I'm encountering arguments that are not addressing the issue. Because of this I think a formal discussion is in order and on the table should be discussions about policies governing the release of cloud images. First, who am I? I'm a Unix enthusiast as well as open source author and contributor. I've been using Debian since 1994 and have continued to use Debian as well as nearly every other Unix that exists to date (even the really mean ones like HP-UX). I was a user of the EC2 Debian/Wheezy AMIs until the rate of changes on the AMIs started ruining my operations. Trying to actually use the AMIs produced though was so painful the best option I could find was to stop using the work of the Debian cloud team because operations are impacted far more by upstream changes to the AMIs than anything going wrong on their own or because of difficulties in EC2 related workflows that the OS is going to solve for me. Its quite clear there is not a fundamental understanding of what my root concern is. One misunderstanding seems to be that I do not want changes to Debian/Stable that are good for commercial organizations. This is false. I do not want changes to Debian/Stable, period. The economic incentives of commercial operations will produce large pressures to make changes right now and is why the topic is run together with "stop changing stable." In the case of GCE the team is co-authoring the image for GCE with Google and this now becomes a direct and extremely important consideration. It also seems that people think that I am against the cloud, against the cloud users, or against Debian trying to help users with the cloud. This is false. I am against a development process targeting brand new technology that supposedly has many new problems that definitely no one knows how to solve well with the results being tested on unsuspecting users looking for Debian stable. Please read that again if it is not clear: the problem is testing the work on users who are expecting Debian Stable not the work, not the cloud, and not the software. It also seems that people think I want to change things to make them simpler. This is false. I want changes to stop. Changes are the problem here. There are going to be some tough questions here and while reading them please ask yourself the following question: whom is the work of the cloud team supposed to be helping? 1) Exactly what is the charter of the cloud team? Why is it here? 2) Is the cloud team producing official Debian images? If not why are the published Debian Wheezy images not clearly labeled as being Debian Wheezy (cloud edition containing experimental software) 3) If the cloud team publishes official images where are policies regarding what is and is not allowed to be published and labeled as Debian Wheezy and/or stable as time goes on? 4) Why do the cloud images contain customizations such as changing the administrative username to admin instead of root? I see two things that are new here. The first is that before "the cloud" Debian Stable came on CDs and there was not any ambiguity about what is and is not Debian Stable. Seasoned veterans of Debian will expect published Debian operating systems to work like they always have. The other new thing is that "the cloud" is the combination of virtual machines and a workflow thats supposed to treat each individual node as being rather unimportant. Virtual machines are easy and well solved in Debian/Wheezy. The concept that each individual OS instance is throw away is a very difficult one to solve and I think everyone doing this work is aware of that. It is so difficult to solve, in fact, that its going to take a very long time to solve, with a lot of testing and a lot of false moves. And that last reason is exactly why developing and testing the cloud solutions on end users expecting Debian Stable is a serious problem. Please compare the output and release cycle of the cloud team against the output and release cycle of release engineering for the CDs. Do I think the cloud team should abandon all the work done to date? No, I think that will be severely disruptive. Do I think the cloud team should not have ever tried to publish software that solves the workflow problems and provide customizations on top of Wheezy beyond extending the CD/DVD image creation process to handle VMs universally? Yes I do. Do I think the cloud team should never have tried to solve it at all? No, I don't. In short the output of the cloud team is not Debian Wheezy yet when it hits the EC2 AMI list it sure looks like it: https://aws.amazon.com/marketplace/pp/B00AA27RK4/ref=srh_res_product_title?ie=UTF8&sr=0-2&qid=139748542126 The only indication here that this isn't identical to an install cd is the tiny warning that you'll have to login is admin and not root. Does this make sense? I hope the cloud team can understand the importance of tightly controlling exactly what is published as Debian Stable. Tyler
Attachment:
signature.asc
Description: Message signed with OpenPGP using GPGMail