On Sat, Dec 4, 2010 at 12:41 AM, Stefan Wallentowitz <stefan.wallentow...@tum.de> wrote: > -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA1 > > Hello out there, > > just my 2 cents on the recent development and my opinion on the future > development. Hello Stefan, thanks for the very well shaped e-mail. I find your observations very reasonable, and will comment below.
> Bahadir, I completely understand your arguments and can agree on the > various trade-offs that arise with business models around open source. I > have just a few points, that I consider relevant: > 1. an important fact in _your_ consideration should be, that codezero > until now failed to build a broader community. With that I mean > developers outside B-Labs, that significantly contributed. Having a look > on the archive of this mailing list, one finds when there was any mail, > it was either "what is going on with codezero" or build issues. The > reason for this may be manyfold and hard to identify, but the project > source, build and docuementation is well shaped. Hell, yeah. SCons + CML2 turned out to be a nightmare for me. I seriously hope that C0 architects had good reasons to chose this build/configuration system in the favor or GNU Make + Kconfig. Mostly because it was new to me, but also it turned out to lack good documentation and examples on net (specially CML2, which files are spread all over the source tree). Apart from the offered targets that build well, any further hacking creates a headache on the build, so I spent most of my time correcting build and configuration files, and not focusing on the OS mechanisms. I would seriously consider adding in-depth documentation on build and configuration process of C0. > 2. Codezero is in a shape it can actually be used in production. As I > understand your currently closed development, the main novelties are in > the userland, right? Hmmm... I did not understand it like this. I guess that multicore support and HyperSwitch trick had to go directly in the kernel. But maybe I am wrong. In any case, I perfectly agree that there is more enough of the open-source code for a serious development aside from B-Labs. > 3. With the decision to change licensing of the sources from now on, it > might still be possible to hope for a community to form and develop the > open branch, while the closed branch is deverging. GPL is not very > friendly to easily get this stuff together (except going back open source). I would opt for BSD licence here (or eCos kind of licence), liberating industry players to keep their design confidential. How will they link their applications or proprietary OSes with GPLed C0 ? Industry likes this kind of liberty, it sounds good in the ears of managers and decision makers. And this still will not prevent OS geeks to play with the code. But even GPL can do the trick (hell, it works for Linux). > Summarizing, you may think about splitting codezero up in the open L4 > kernel part and closed userspace stuff. With the kernel you can hope to > gain attraction from developers, that are really willing to bring > codezero forward, while a business model around the userland, boards, > configurations, support might work out for you. And here also you can consider eCos model and see eCosCentric's business model : http://www.ecoscentric.com/index.shtml. > It will be hard to find the exact part that stays open and which is > close. Having closed stuff in the kernel in your house, makes sync'ing > stuff hard in case there is a developer boost coming up. Nevertheless > you can bring stuff to the open source version at any time. This is also why you should consider changing GPL licence. AFAIK, GPL will prevent even you to link kernel with your closed proprietary parts... Rather find a licencing model for C0 kernel that will allow you to have kernel parts, board configurations, drivers, userland. As I said, BSD or eCos similar. > I would suggest, to try to establish a simple open source project > management such as Trac with the current codezero repository and see > what happens. Yes, yes, yes ! This is very important observation, and I would to underline it. And it should be rally easy to do. It gives you a great and easy editable documentation placeholder, all along with git interface and wiki pages. Easy to put pictures and figures (dunno about videos, but I guess). As an example I'll give OsmocomBB (http://bb.osmocom.org/trac/) or SoClib (http://www.soclib.fr/trac/dev) families of project, a very useful and fast-blown presentation. Or you can consider setting a wiki page like, for example MediaWiki, like M5 (http://www.m5sim.org/), although DokuWiki (http://www.dokuwiki.org/) seems to be even better for development purposes. Personally I prefer Trac because of source browsing option via git and svn interfaces. > The current status should be sufficient to attract people, > maybe try to advertise it. This is something I also underlined in the previous e-mail : attract the hackers by advertising your project. Project is great, challenging and innovative by itself, all you have to do is to work more on informing people that it exist and what possibilities it offers. And also, create better development infrastructure we mentioned (Trac, build documentation, etc...). > Meanwhile you can try to build your business > model with clear boundaries between open and closed source and bring > your stuff in, participate in the open source project and try to make > money out of your really nice novel stuff. I guess something like that was Bahadirs idea from the beginning, although contract signing procedure sounded really weird to me. I personally do not like putting my signature on any contract I do not have to, so I sent Bahadir e-mail with a request to explain me what is it about. Is this kind of procedure really necessary (as it is quite unusual and can be repelling for a hackers looking for a next challenging project to dive in). What are real boundaries here between closed and open source ? I think all of this is not so clear and has to be re-defined, as Stefan mentioned (maybe even re-licenced to some licence other than GPL). I hope that this comments and observations will help making this interesting project more attractive and advanced. Kind regards, Drasko _______________________________________________ codezero-devel mailing list codezero-devel@lists.l4dev.org http://lists.l4dev.org/mailman/listinfo/codezero-devel_lists.l4dev.org
