David Luff writes: > As for 1.0, although its just a number, I personally think its a > pretty significant number, and probably worth a bit of work > polishing bugs , user interface, and installation problems out as > much as possible before release.
David, Definitely we want to get out releases that have no major bugs and have all installataion issues resolved. However, there are a couple things that work against us. 1) Time. Putting out a good release takes an immense amount of time. I figured that I had at least 40 hours of real time into the 0.8.0/0.9.0 release. That's over and above my real job, my family, etc. etc. People need to realize how much time, and how much effort putting out a good quality release actually is. People might argue that there were still a lot of problems with the 0.8.0/0.9.0 release which goes to my point exactly. I really needed to find 80 or 120 hours or more to do things right. 2) There seems to be a principle at work that very few people download and test development and pre-releases. Mostly it's a few developers who already know all the tricks, already have all the prerequisites on their systems, etc. This means that the big flood will come after 1.0 is officially released. That will be when all the bugs and installation issues and other problems are reported, and unfortunately not before. I'm open to suggestions, but I've tried a lot of things already, and what ever I do has to be able to fit into my reasonable time limitations. Finding bugs after a release isn't necessarily bad in the grand scheme of long term flightgear development, but it can make it look like we do a careless job, or ignore certain platforms or compilers. I can vouch for Debian Linux since that's what I run personally, but I don't have time or resources to build and test on other platforms myself. This probably goes to David's point that CVS is actually the most stable, becuase it's not until after we do a major release that the bulk of the real bug reports and usability issues start pouring in. However, I'm usually not keen on another 40-80 hour go around right after finishing the last release. 3) Expectations are somewhat different for us than many other open-source applications like "autoconf/automake". Those guys just wrap up a tarball and release it and are done. We have to coordinate with the base package release, people expect prebuilt binaries, cd-rom images, etc. I'd love to just do a source code release, it would make my life simpler. Maybe I should consider it seriously. But then inevitably, I personally (since I'm the primary flightgear contact) get flooded with questions, complaints, people howling that they shouldn't be expected to compile an application from scratch, or have to learn unix, or have to read instructions, or type from a command line, etc. etc. People want a Windows/Mac/Linux installer. Download one big file, double click on the icon and it's installed and all perfectly native to whichever platform they are on. I haven't the time to do this myself, and apparently (since we don't have these sorts of things) no one else does either. But there seems to be an underlying expectation that someone should be doing this stuff. I'm not sure who that would be though. 4) There is too much of an attitude or expectation overall that some magical core of developer(s) (i.e. someone else) will do everything, and make everything work, so that any end user can then point and click and everything works perfectly the first time. But, none, or very few end users are willing to try the development and prereleases and report bugs in advance. A few "other people" are expected to keep all the documenation perfectly in sync with development. "Others" are supposed to make sure everything works perfectly with your platform, etc. etc. I think our limited attempts to do this draw people into thinking that "others" have infinite time and should just make it all work magically. But this is open source, we are the "other" people that need to make sure this happens. I can do a bit of it, David can do a bit, Norman, Erik, Christian, Andy, etc. etc. (not to single out specific people). But when FlightGear-1.0 doesn't build cleanly on (for instance) Mandrake version X.Y.Z using gcc-x.y.z who's to blame? The person encountering the problem needs to recognize that the developers have done their best and already put in way more time than they should have. We need a lot more help in these sorts of areas than we are getting. I'm not really ranting against end users here. I understand that a lot of them really don't have the tools or experience or resources to make useful bug reports or participate in fixing problems themselves. But when they do run into problems, they need to have a bit of an understanding of what to expect in terms of who might fix them problem and how soon. We really are trying to make this as easy on people as possible and accessible to the masses, but to do so is an incredible amount of work. Much more work than one or two or three people can do in their spare time. Long ago I gave up striving for perfection. Instead I strive to do the best I can do in the time I have. If we want to do better than that, we need a lot more help from a lot more people. > It might also make a good opportunity to test Curt's contention that > the front page of Slashdot wouldn't break his servers :-) I say bring it on. :-) > There seem to be some problems with the JSBSim gear model which I'd > have down as a show-stopper for 1.0, I haven't heard this discussed. You should probably take this up with Jon/Tony on the JSBSim list. > and I'd have thought that displaced thesholds and the arrows > pointing to them would have to be pretty high on the list of > features that would be expected to make it in. Do we actually have these in our airport data? If so (or if the data could be added) I'd be willing to work on it. The airport code is still relatively fresh in my mind. > My personal hope as a non-US citizen is that world-wide DEM-3 data > from STRM becomes available prior to 1.0, but I'm not holding my > breath on that one any more. I grabbed the 30 meter (1 arcsec) USA data, but I haven't had time to start looking at building scenery from it. That's another can of worms ... there is a ton of things in terragear that I'd like to go through and rework and improve ... someday ... Regards, Curt. -- Curtis Olson IVLab / HumanFIRST Program FlightGear Project Twin Cities [EMAIL PROTECTED] [EMAIL PROTECTED] Minnesota http://www.menet.umn.edu/~curt http://www.flightgear.org _______________________________________________ Flightgear-devel mailing list [EMAIL PROTECTED] http://mail.flightgear.org/mailman/listinfo/flightgear-devel
