A few comments:
but if the problem can be well defined, then it may become irresistible to someone who can fix it
So far, most problems have been fixed by those who can, when the problem becomes a personal one. If the problem isn't a problem for those who can, it won't get fixed unless somebody pays them to fix it. Even then, those who can are so busy with other things, that they don't want to take on more projects.
I also appreciate that USB has already peaked and interacting with the cloud is the future.
For many years, IBM pushed for "diskless" computers. They failed to accomplish that, simply because most people realized that storing data on somebody else's computer was a security problem, at best. The "cloud" is simply a new version of the same thing, with a few new twists. I don't see it being as useful as many seem to think it will be. One high profile security breach, and it will go away. It is only a matter of time.
USB may have a new life, now that USB 3 is here (and we desperately need a driver for that), but I expect that USB 3 will also run out of steam in a few years. It is just too slow. ESATA is the best alternative, followed by NAS storage (with at least a 1Gb/s interface). Of course, something else may come along to replace all of them, and then we will need support for that.
1. Some devices are large floppies, and have no partition table. We should recognize them as partitions anyway so we can use our FAT32 IFS.
This is true, but it is just plain lazy (on the part of the manufacturers) to not have proper partitions on something larger than 2 GB (that probably won't change). It would also be better to be able to use NTFS, rather than FAT32, but nobody seems to want to improve the NTFS driver. FAT32 is also not very good, but it is the only real alternative for sharing data between OS/2 and Windows.
2. LVM should recognize non-LVM volumes and give them a drive letter or mount them in some way.
One possible solution is to put a driver (filter?) between the USB driver, and the LVM stuff, that will add LVM information on the fly, when it doesn't exist. How that can be done, I leave to those who know.
In fact, LVM already does that, with small volumes. The problem lies with recognizing large volumes. When IBM did the LVM support, they made it far to ridged, and it simply doesn't allow for volumes that don't follow the "rules". Of course, nobody else thought that following the "rules" was necessary (they probably don't even know that "rules" exist), so we end up with the problems that we see today. It may be possible to patch the current software to handle the problems, but then we run into the copyright/patent problems.
3. The system should be more flexible with devices with incorrect geometries. If it works OK with Windows, Linux and Mac, it should work for us, too.
This too is caused by IBM trying to stay within a ridged framework, while nobody else bothered to do that. The other cause of our problems, is that IBM didn't plan for larger media, so the drivers etc. are not capable of handling it. Since we really don't have proper access to the driver source, and we have very few people who understand such things, it is very much hit and miss to fix the problems. Hopefully, the AHCI project will fix the internal hard disk support, and the USB driver project will result in USB 3 support, but it is slow going because those who work on such things are far too busy with other things (some more important, and others not so important), plus they need to learn how to do it as they go along. Again, an additional driver (filter?) to adjust the data so that OS/2 can use it, may be the answer.
One of the most depressing parts of all of his, is that there is, at least, one guy who is probably capable of fixing such basic things, but he is off on a mission to make games work better. Personally, I don't think that games are going to be very important if the base system won't work.
Now, for
What do we need?
What we need, is for IBM to release the source for OS/2 (probably won't happen). Then, we need hardware manufacturers to be more open about what their devices need to make them work, so we can create new drivers (probably won't happen). Then, we need more people who are capable of programming device drivers, and other parts of the system, and some coordination to get the important parts done (probably won't happen, although it seems that Mensys is trying). We also need a few thousand competent programmers, who are willing to donate their time (probably won't happen).
One thing that can be done, is for more people to step up and help those who can do something about these problems. Much testing of new software needs to be done, and all it takes is a bit of education about how to do that, and how to make proper problem reports, and many more people can help out. Complaining in the forums is not helpful, although venting some frustration probably doesn't hurt the individual who does it.