Werner Almesberger wrote: > Mike (mwester) wrote: >> I rather agree with Paul on this. The proof-of-concept has already >> been done, and all the ingredients are pretty much pre-packaged already. > > True, the basics have been around for a very long time. What seems > to be missing is something people can try on their GTA02 and then > go "now if it could also do this ..." (and then implement their > extension).
You may be right. There's no doubt that there are many more "application-level" developers in the community than "system-level" developers (substitute the correct terms, if such exist). So where *have* all the developers who could do this gone to, then? :( >> All the bitbake recipes, including kexec-tools as well as a mechanism >> to create a minimal rootfs with a menu are already in OE, and are >> running on other devices. > > I'd rather see the actual merge happen after OE. OE is a great > heavy-duty tool for the momentous task of building distributions, > but our attempts to sell it to people as a build tool for their > day to day work has been an abysmal failure. Well, that's an opinion. OE may be difficult, and may pose a rather high barrier-to-entry -- but the IRC channels and mailing lists are full of people who have successfully scaled that hurdle. The reason I suggested OE is simply because most of the distros out there use OE as the build tool in the first place, and the generation of the minimal rootfs for a kexec-boot would amount to a few extra minutes of build time, since the overwhelming majority of the work has already been done by the time the distro is built. But no matter - any tools will do for this. I've put together a prototype simply by hacking an existing JFFS2 image with a little scripting added. So OE, or not, the tools are all there - tested and waiting. I'll be happy to assist anyone who wants to do this, no matter what tools they wish to use. (I, unfortunately, have been rather short on time over the last month or more, and can't really do much more with what I've got.) > We now have convenient tools for putting together things from > pre-compiled packages so OE can do its work in the background. > This greatly reduces the entrance barrier. (Again, the same > theme - make it easy for people to contribute and they will.) Yep - putting the toolchain and SDK together was an excellent bit of work that got fed back into OE from Openmoko. With the toolchain and SDK, I think that the biggest barrier remaining right now is the fragmentation of the distros. I wonder how many folks are thinking "I have a real cool idea, but I think I'll wait until one of the top four or five distros emerges as a winner and stabilizes a bit". >> Perhaps when the GTA03 comes out without u-boot at all, there will be >> renewed interest in a menu, and this will come to happen. > > Wouldn't you hope that GTA03 will be a bit less of a "matures at > the customer's premises" kind of product than its predecessors ;-) Originally, yes. But having been here since the early days of the GTA01, my expectations have changed somewhat. ;-) But hey, look at it this way: it's easier to surprise me this way. > > - Werner Mike (mwester)
