On 11/30/2013 10:31 AM, Joerg Stephan wrote:
Am 30.11.2013 13:38, schrieb Robert Schweikert:
When i talk to developers, they dont want to switch to newer upstream versions every 18 months. No application lifecycle can handle this.
But the question here is, are applications that we do not build in OBS part of our target? If those applications are open source, they possibly should be part of our target, then why are they not in OBS? What does it take to get those companies/developers to build their application in OBS?
Well, i dont get the question! There are several things around which depends on software we provide. Any Apache,Nginx,PHP,Perl,Python,ruby Software can be developed and operated on our distribution, but maybe deprecated or needs review or redesign 18 months later.
Could be, but there is no solution to this problem. Even in the enterprise distribution world where release cycles are excruciatingly long we still get complains that things change too fast. At the same time of course we hear people tell us that we do not move fast enough. There will always be that part of the spectrum that wants "no change" and then there are those that want "change everyday" and we'll have everything in between. Therefore, I do not believe there is som magic number that creates a magic alignment of the starts such that more users would be interested in running openSUSE. Not in the current paradigm where people look to releases as being the basis of their operation.
While it takes several months to migrate it. Thats the point, i guess not that any company running a shop or a website will put it on obs, but maybe the want to run it on opensuse.
Yes and no. Yes, because the company that just runs the web site and gets the code from someone else is certainly not interested in the churn if there is a potential of the app they are running being broken. No, because the company that just runs the website should not have the concern of the app being broken. If the web application or other application were integrated in OBS, and the devs can certainly run their own instance without divulging their code to the public, than the developers deal with generally incremental changes rather than a big clump of changes at release time. Thus, at the end user the churn is not really visible becuse the app is not broken in the first place.
I dont talk about software itself, i talk about companys/user using software to run websites, office programs etc.
Once open source applications build in OBS the life cycle discussion goes away, for developers; and to a certain degree for users as well. For developers the app is build in OBS and is always integrated with the "latest and greatest", thus handling change, which is incremental but frequent becomes a small effort compared to handling the accumulated set of changes at every release. For users the life cycle becomes somewhat immaterial as the application has already been tested on the release as it was part of the development process.
When it comes to proprietary applications I am not certain we should care. I do plenty of that w.r.t. my $DAYJOB and let me tell you, it's not pretty and no life cycle is slow enough for proprietary app developers not even the SLES life cycle of almost no changes for eternity.
No, but 5 years support and another 2 years if you pay for it. Well, i can tell you from my $DAYJOB that companys care about that, the longer support the better.
Yes, that's why the enterprise linux business exists. People that cannot deal with changes pay to have old stuff maintained for eternity. And it is still cheaper than paying for AIX, Solaris, or HPUX.
Upgrading an operatingsystems costs a lot of time, which means a lot of money. Administration who runs the updates and developers who tests it, even if everything works fine it takes weeks.
Yes, I know.
Well, that is a bit too generic for my taste. openSUSE is not a solution for a company that needs proprietary applications. But than again neither is almost any other community distribution. One of the key ingredients here is commercial software support. Most companies will rely on some sort of proprietary applications and those generally do not support community distributions. For companies that do without proprietary apps or are willing to put up with a mixed environment I do not see the life cycle as a disadvantage. No one forces anyone to upgrade. Ubuntu LTS, that has been mentioned as an example, has a 24 month life cycle, we have 18, or 36 with Evergreen, thus I'd say we have the time frame covered and I refuse to believe that 2 years is some magic sweet spot that makes things work. I think the 2 year period is a number that was pulled out of somebodies.....
So we come to the point. So tell me, and if you read the other mails, what is openSUSE a solution for in your mind?
From my point of view openSUSE is a great development platform, unless one is a Java developer. It is solid, the pace of change is reasonable, upgrades from one version to the next mostly work without hiccups and with OBS I can pull in the latest an greatest in my area of interest, Python, Perl, etc. I have little insight into running servers for other people or being an admin of 50+ machine installs. However, if somone is running such installs on a distribution that is community supported, i.e. best effort with no one to blame than they are probably skilled enough to deal with the cycle we have. I don't see how adding 6 or 12 month make a difference. Again, there are people that complain that SLES moves too fast, thus increasing the length does not appear to be the answer.
Thats the question, thats the point you replied to earlier. What is it, whats the target user, what is this all about. Where do we stand and where do we wanne go. Servers, business, gamers, grand mothers who wants to write an email?
Grannies and email writers will eventually end up on Chromebooks, I think. The next time I am at my mothers I will probably move here to a Chromebook. Her computer is approaching 8 years and things are bound to fail before too long. But for the past 8 years she's been running openSUSE with upgrades on the 18 or so month cycle with no issues. Why switch? Simply for the reason that she doesn't need what we provide and the Chrombooks are a chunk cheaper than a laptop that does not carry the Windoze tax. Yes I vote with my wallet and have not bought a machine with a Windoze tax connected to it in almost 10 years. Every other category is probably fair game. However, all this categorization always makes me wonder why people are so interested in creating these artificial barriers and fences. That's not what Linux is about. If the kernel folks were only interested in running on servers we wouldn't have Android, Tizen, Automotive Linux etc. The beauty of it all is that it is adaptable and flexible. If we have people in the community that want to create a openSUSE-Server with 5 year "support" distribution let them do it, why are we even "arguing" about that? The point here is, and I suppose this is discussed in the related threads as well, that until now we have run the same release mill. This mill has required a lot of effort by the openSUSE team as they carry a lot of the final release burden. It is appreciated and has helped us to knock out some great releases. That team appears to be interested in trying something different, which is their prerogative. What the rest of us have to answer is if we have the means to step up and continue the cycle, if we want, if we need to ask the openSUSE Team to help with this cycle for another release or two until the community is fully prepared to shoulder the 8 month release cycle or if we even care enough to do releases. We do not all have to march to the same drum as Evergreen and Tumbleweed have already proven. It is a community effort and trademark approvals these days go through the board. Thus if there is a group of people that wants to create openSUSE-Server and release this as a distro I personally see no problem with that. Why would anyone stand in the way of such an effort? Maybe going forward there will no longer be "the one and only" openSUSE distribution, frankly with Tumbleweed and Evergreen that doesn't exist anymore already anyway.
With the life cycle you might have a problem as that has a very wide ranging effect. However, if you'd like to see an openSUSE-server "distro" subset, there is nothing in your way to start the effort and get working on it today. The tools are there. This is nothing that can/will.should be decided by discussions or some "magic power group". Having a server profile or subset is a matter of someone doing the work and pushing it into factory and having a team of contributors from around the effort. Be free, go and do the work.
Good answer to any upcoming question. "Go and do it yourself". It should be discussed by the community and at least by this mailinglist, if the discussion says we dont need it, why should i do it?
This is not a dictatorship. Just like things that were introduced in the kernel for mobile stuff turned out to be useful for the server eventually and vice versa (with multi core stuff showing up on mobile) there will be things that an openSUSE-Server team with a 5 year "support" plan will bring back to the community at large. That's how organic growth works.
Lets make a deal, board and project ML should discuss it, if we come to the conclusion we want a server subset, i will be the first one in the team and the last one who leaves it.
Well in this thread alone I'd say there were already 3 different people that said we need a longer more server oriented life cycle. I think that's a good start to get a server only team going. We have examples of this stuff right here in front of us. When the KDE team moved to develop and maintain KDE4 Illya decided he wanted to keep KDE3. So he did, or is doing the work, and I suppose there are users. I maintain OpenNebula repositories and we have openSUSE users that use OpenNebula instead of OpenStack. If I really wanted to I could probably get the OpenNebula packages into the main distro to live alongside the OpenStack packages. In most cases it really only takes someone or a few people to do the work. The answer of "do the work" is not a cop out, it took me a while to come to that conclusion when I first got involved oh so many years ago in open source. But as John "Maddog" Hall says, don't wait for a "letter of permission", it will not arrive.
Well, this is one of my problems, who is somebody? While I agree with Stephan, to a certain extend, that our strategy discussion was somewhat of a disaster, we did emerge with a document. This does describe the things a large part of the community agreed upon as our goals. I do not see where yet another list of goals created by the mystical "somebody" will make any difference.
Well, i hoped someone from the board, if i got it right http://en.opensuse.org/openSUSE:Board seems to be part of the description. According to our speech it seems the mystical "somebody" will never be me and never be you.
This puts me in a bit of a difficult position, thus let me make certain there are no misunderstanding. The following paragraph is my opinion and I am NOT putting words in the mouth of my fellow board members. Correct, the board is not here to set technical direction, release cycle, target audience for the distribution etc. Those decisions need to come from within the community and if parts of the community decide that a 5 year release cycle of openSUSE server is the right way to go and others decide that the 8 month cycle is just fine than what can the board do? Forbid a group of volunteers to contribute to the project? I think not, that's certainly not going to go over very well. As Kostas said, people here volunteer their time and efforts. Presuming that the board could tell people what to do and how to spend their volunteer time is at least in my opinion over reaching. That we had 1 primary distribution for a long time is great but that does not necessarily imply it has to stay that way going forward. Later, Robert -- Robert Schweikert MAY THE SOURCE BE WITH YOU SUSE-IBM Software Integration Center LINUX Tech Lead Public Cloud Architect rjschwei@suse.com rschweik@ca.ibm.com 781-464-8147 -- To unsubscribe, e-mail: opensuse-project+unsubscribe@opensuse.org To contact the owner, email: opensuse-project+owner@opensuse.org