Ed, before I say anything else let me just say that I'm much more comfortable with this after reading these last two messages. I still have some concerns, but unfortunately limited time to formulate/express them. Still, you've given me comment bait:
Ed Leafe wrote: > Versioning and distribution has always been a very big shortcoming > of Dabo. It is by far the #1 source of confusion to newcomers to the > framework. I monitor the Python and wxPython lists, and I can't tell > you how many times I read "Dabo looks great, but I couldn't get it to > work". Well, did they ever ask us for help? Such messages should be ignored, IMO, as they don't add anything beneficial to any discussion. And honestly, I've been monitoring python-list and wxPython-users for years, too, and don't remember seeing more than a single handful of such messages. > Our releases have also been very spotty. It takes a lot of time to > organize, test, and put together an official release. As a result, The time it takes to organize a release is basically testing, testing, testing, and summarizing what has changed. IOW, the releases aren't made lightly, and we are reasonably sure we know what the current problems are and that dabo is working smoothly when we release. > we've gone over half a year without a release at times, and as a > result many potential users who go with released versions have code > that either is buggy, won't work, or won't do what people are > discussing on the lists. Buggy: means that we failed to backport our bugfixes. Mea culpa for much of that, as it is time consuming to sort out new stuff from bugfixes, and to remember to backport the relevant bits. Won't work: means that we shouldn't have released it in the first place? Won't do what lists are discussing: I'm not sure I can come up with an example, here, but I think I know what you are referring to. Hold that thought, though. In dabo-users, I'd expect discussion related to the most recent released stable version, and to support that. In dabo-dev, I expect discussion related to the current trunk. Ok, bring back that thought. I believe the main thing that has bitten you in the past is that your (excellent) screencasts have focused on what is happening *right now* in the development of the class designer. IOW, the trunk, for developers. Then, you show the screencasts to dabo-users, and they get confused because their versions don't match what they see in the screencasts. Understandably, people are confused and frustrated. But the error was in releasing a screencast of the development branch to people using the released, stable branch. > Our current distribution system is actually a bunch of separate > systems: > > - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - > 1) Download the official released versions, and run setup.py to install. > Pros: simple and direct > Cons: usually out-of-date by the time it is installed. Difficult to > upgrade. Stable releases are, by definition, out of date, as development isn't happening in the stable branch. This isn't a con, but people need to understand they are installing a codebase that isn't going to change much. Difficult to upgrade? 'python setup.py install' was simple and direct for the initial install, but difficult for the upgrade? > 2) Nightly Tarballs. > Pros: Easier to keep current > Cons: requires manual updating. Can sometimes result in getting a > non-working snapshot that will not be fixed until at least the > following day. That con is huge. I've never really been happy with the availability of the nightly tarballs for this reason. > 3) Subversion. > Pros: latest and greatest copy of the framework. > Cons. One more thing to install and manage. Can get changes that are > broken. Subversion is for developers, not users, and with it you can get any released stable version as well as the active trunk. My take: Pros: Most flexibility; easiest for developers; easiest to keep current. Cons: Not for users, just developers. > - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - > > I'd like to simplify this. I see two primary distribution models: > > 1) Subversion: intended for those who are actively developing the > framework Yes > 2) Web updates: intended for those who are using the framework to > develop their apps, or who are exploring/learning about Dabo. I'm still going to hold judgment on this, because I still firmly believe that for the users of the framework, the stable releases *will be* the best and recommended route for them. I say "will be" because I recognize there have been problems recommending the stable releases, mostly because we don't have time to properly support them while developing the new stuff. Now, an update mechanism that would be able to say "oh, look, we are at 0.9.3, and 0.9.4 is now available" would be a great thing for this type of user, and I'd support that 100%. But, in your other message you gave as a scenario that the user downloads a stable release and installs it (let's say 0.9.3). What I'm confused about is that the stable branch is different (older) than the trunk, so when we mark a trunk revision as 'stable enough to easy update', now all of a sudden does the user effectively go from stable (0.9.3) to trunk (0.10a), receiving all the new stuff that has been added or changed in trunk? If so, I don't think I'll be able to get behind this idea, because it violates the separation between the development and stable branch. Please forgive me if I haven't explained myself well, or have misunderstood the point of easy update. > These divisions parallel the distinctions between the dabo-dev and > dabo-users lists, and I think will make it simpler for anyone using > Dabo to stay current without getting too close to the bleeding edge > unless they need to be. I'll say it again: regular users using the trunk is too close to the bleeding edge. They *should* be using the most recent stable version. The problem is what you elucidated before: our releases and backports are spotty, and what I observed: we talk about the trunk branch to the users before they have the opportunity to play with the new features. In summary, I like the idea of easy update, but I fear my idea for it will blow away your intention for it (limit it to updating just the released point versions). -- pkm ~ http://paulmcnett.com _______________________________________________ Post Messages to: [email protected] Subscription Maintenance: http://leafe.com/mailman/listinfo/dabo-dev Searchable Archives: http://leafe.com/archives/search/dabo-dev This message: http://leafe.com/archives/byMID/dabo-dev/[EMAIL PROTECTED]
