Announcement

Collapse
No announcement yet.

Fedora Might Shift To A Tick-Tock Release Cadence

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • #21
    Originally posted by AdamW View Post
    ...you don't provide any particular argument as to why it would make sense for us to do the thing that makes you happy as opposed to the thing that makes other people happy...
    I don't mean this contentiously, but does the Fedora Project periodically investigate if sufficient demand is there, and, if so, take that into account? I have personally enjoyed Fedora 20's extended primacy (i.e. duration as the go-to release), but I realize the Fedora Project has no obligation to me. I don't think having a designated release with a slightly longer life span poses a threat to RHEL, nor is RHEL a perfect substitute in this regard, although it seems that whenever this discussion comes up, invariably it is dismissed with "use RHEL". I believe there is a Goldilocks zone between the life cycle of a conventional Fedora release and the life cylce of a RHEL release, but perhaps I am a lone voice in the wilderness. I understand if the resources aren't there to manage both LTS releases and interim releases, while still maintaining Fedora-ness.

    Also, while I'm all for release-when-ready, predictable release cadence does factor into some organizations' decisions as to which distribution to use. Would the tick-tock model pretty well guarantee a release within a month's span every year, whether that be May & October of each year, or some other schedule?
    Last edited by eidolon; 06 December 2014, 03:11 PM.

    Comment


    • #22
      Originally posted by eidolon View Post
      I don't mean this contentiously, but does the Fedora Project periodically investigate if sufficient demand is there, and, if so, take that into account? I have personally enjoyed Fedora 20's extended primacy (i.e. duration as the go-to release), but I realize the Fedora Project has no obligation to me. I don't think having a designated release with a slightly longer life span poses a threat to RHEL, nor is RHEL a perfect substitute in this regard, although it seems that whenever this discussion comes up, invariably it is dismissed with "use RHEL". I believe there is a Goldilocks zone between the life cycle of a Fedora release and the life cylce of a RHEL release, but perhaps I am a lone voice in the wilderness. I understand if the resources aren't there to manage both LTS releases and interim releases, while still maintaining Fedora-ness.

      Also, while I'm all for release-when-ready, predictable release cadence does factor into some organizations' decisions as to which distribution to use. Would the tick-tock model pretty well guarantee a release within a month's span every year, whether that be May & October of each year, or some other schedule?
      I'd say it's more 'ongoing' than regular, but yes, absolutely, this kind 'what's the target market' / 'what's the right strategy' debate goes on allllllll the time (the FESCo, Board/Council, and devel@ archives are littered with them, if you really want to spend your weekend reading some :>). The Board spent a couple years on a fairly concerted effort to come up with a coherent approach.

      I'm not saying the question is invalid or Fedora definitely currently has the right answer, just saying that I think you have to provide a bit better justification than 'this happens to be the model that would satisfy *me* personally' in order to convince a given distribution that it should adopt a particular model. The distribution needs to be convinced that it's the right model for the distribution's target audience - and of course, the target audience debate is open as well. But you do need to engage with the whole question, not just 'this is what I want'. Well, I mean, if you're clear that all you're doing is saying "hey, this is my personal opinion, throw it in the pot of data", then 'this is what I want' is fine. I guess it's 'this is what I want *and therefore this is certainly what must happen!*' that gets my goat.

      There's clearly a gap (probably several gaps) between Fedora's approach and RHEL's. Maybe not everyone Fedora/RH-ish would agree with me, but *personally* I've got no problem saying, hey, there are several other distros that operate in that gap. If you want something in it, maybe one of those would work. I'm not really sure any one distro / distributor needs to conquer the world - not sure the RH/Fedora conglomerate necessarily *needs* to have a distro in that space. We don't need to corner the marketplace. =)

      I don't think it's at all unreasonable to consider the question of whether Fedora could move its approach slightly. Personally I think anything along the lines of 3-year release cycles/support lifetimes is a bit too radical a change and would more or less result in an entirely different project, it wouldn't *be* Fedora in any meaningful way any more. But that's just one monkey's opinion, of course. I'm not on any of the boards that decide these important questions, I just sit here bashing stuff till it breaks.

      Entirely personally I'd kind of be interested in either a rolling cycle or something more along the lines of 9-12 month cycles, but as I said we've had these debates before, and there are absolutely good points to be made against either of those approaches. It's a complex area, and really I just have a knee-jerk reaction against people who are more or less saying "well OBVIOUSLY this answer is correct and if you don't see that you're an IDIOT". I like nuance, myself.

      Comment


      • #23
        Originally posted by AdamW View Post
        ...I'd kind of be interested in either a rolling cycle or something more along the lines of 9-12 month cycles...
        Given their respective foci, might the three product groups (cloud, server, workstation) have divergent wants in this regard?

        A stable rolling approach could be quite appealing, but I have the impression that would be a harder sell to the shot callers than 9-12 month cycles.

        Comment


        • #24
          Originally posted by eidolon View Post
          Given their respective foci, might the three product groups (cloud, server, workstation) have divergent wants in this regard?

          A stable rolling approach could be quite appealing, but I have the impression that would be a harder sell to the shot callers than 9-12 month cycles.
          In theory, making that possible was part of the motivation for the split, yeah - Server for instance may want to look at releasing less often, Cloud at releasing more often.

          In practice it wouldn't be entirely straightforward. We have quite a lot of tooling and process built really strongly around there being one release of 'Fedora'. If you started from scratch to build what the Fedora "Products" are theoretically supposed to be, you wouldn't build the infrastructure Fedora has. To make that possible would require quite a few changes to scheduling, development and QA policies, and to release engineering tooling especially.

          Comment


          • #25
            Originally posted by AdamW View Post
            So what *is* your point exactly? Yes, we decided to take some time to make some major changes to Fedora. It was announced and planned long in advance. This is factually true. OK. What point is it that you're making, exactly?
            Well, it's simple. The post I was originally responding to was asking what was broken, that the Fedora release process needs changing. To which I pointed out that the way F21 had been handled showed that Fedora were clearly finding that the traditional bi-annual release model wasn't quite working for them.

            Please don't take this as an attack on the project, as it seems from your defensive tone - I'm actually a happy Fedora user. I was simply replying to that question of "what was broken?"...

            Comment


            • #26
              Originally posted by Delgarde View Post
              Well, it's simple. The post I was originally responding to was asking what was broken, that the Fedora release process needs changing. To which I pointed out that the way F21 had been handled showed that Fedora were clearly finding that the traditional bi-annual release model wasn't quite working for them.

              Please don't take this as an attack on the project, as it seems from your defensive tone - I'm actually a happy Fedora user. I was simply replying to that question of "what was broken?"...
              No, I just wanted to know what your point was, because I couldn't work it out from your prior posts.

              I'm afraid it still doesn't seem valid to me. Deciding to use a longer cycle as an exception in order to get some specific larger-than-usual changes done doesn't seem to me to mean that "Fedora were clearly finding that the traditional bi-annual release model wasn't quite working for them."

              Comment

              Working...
              X