"Uros Nedic" <urosn at live.com> on August 08, 2009 4:54 PM:
> My basic thoughts was about how to get bigger market share > in huge market of PCs/laptops. Only solution I could see to > be there is to have more high quality ported applications which > in turn will bring us more users. More users will, in turn attract > more companies to port their applications to OpenSolaris. Lets > call it positive feedback. Well, the first step has been achieved: Toshiba sells OpenSolaris laptops. (As for how aggressively they are marketing them, I can't say...) I think the next step is to somehow become interesting to a low-cost netbook manufacturer. For example, the Eee running Linux certainly really helped Linux's visibility as a viable desktop for "average users". Providing a *full* OpenSolaris install (more than just the base software, but rather something tailored for actual use without having to tell the user to use the package manager to do basic "desktop"/"web" stuff...) on a netbook around the $99 price point would certainly make it visible. Netbooks are nearing this price point in the USA right now (I keep getting emails for EeePCs running Linux at $129.99). Does anyone have any contacts with netbook manufacturers (ideally, one just entering the market!) which could potentially make this a reality? > If we do not achieve it, then nobody would like to buy OpenSolaris > PS/laptop just for running applications, which in turn nobody would > like to port their application to OpenSolaris, which again in turn > will lower OpenSolaris users, etc. Lets call it negative feedback. For the most part, little "porting" is needed these days. :) > As far as I could see we are now in state of negative feedback, > and the main problem could be formulated as - how to make transition > from negative feedback to positive feedback. In my opinion, the lackluster approval rate on SourceJuicer doesn't help people port software which would then be available in /contrib. This certainly cannot help. Also, in my opinion, /contrib needs to be a default repository available to users. The community really is one of the best resources for porting software. > One of ideas was to start VirtualBox and inside start Windows OS, > and then run desired application. All is legal but I saw so much > overhead in this idea. It is also expensive. :) Wine actually does a pretty good job with most applications. Day-to-day, I run VirtualBox for only 3 Windows applications. Everything else runs under Wine and works "natively" as far as I care. :) > Next idea was that inside VirtualBox we start Mac OS X and > then run our favorite application. That was better idea regarding > overhead and speed but also it had two negative side effects - > it is not possible to do it legally and we still have to run > another OS inside OpenSolaris. It is potentially possible to do legally, but the kernel (Darwin) which OS X is based upon is not yet supported by VirtualBox, and it's unclear if/when it will be. The PureDarwin distribution of Darwin has a pretty good write-up of the issue, while omitting the fact that if the VirtualBox guys get Darwin open-source distributions running, they'll be pressured by users and probably generally attacked with FUD when OS X still doesn't boot. So, I would say that this approach is infeasible for the foreseeable future. > In past few weeks we could saw announcement from company called > Bordeaux Group who 'ported' some applications from Adobe and Microsoft > to OpenSolaris via WINE. It was great since we do not have to start > guest OS and in turn we have much lower overhead and faster execution. Wine's quite mature on Solaris. :) I've actually not looked very hard at Bordeaux, as it appears to include patches which aren't in mainline Wine, which then make the Wine unsupportable via the normal channels. I'm actually trying to get rid of the few remaining patches in the spec-files-extra SFEwine right now. ;) Boomer + Wine is 100% supported by upstream Wine, by the way. :) And I've been trying to submit patches to winetricks (a third-party script of things which help Wine "work" better) to make it more happy with Solaris. Hopefully, the last batch of patches I submitted against winetricks should eliminate the remaining portability issues when they make it into SVN. I think Bordeaux relies heavily on winetricks, based on their site. (With Bordeaux, my impression is that you're paying for "spit and polish", support, and people to watch the list for patches that aren't yet main-lined for whatever reason. Not a bad business model, and it suits Codeweavers fine. In fact, I use Codeweavers Crossover on OS X rather than "real" Wine a lot of the time due to OS X not being a first-class Wine citizen yet, in my opinion.) > My proposal relies on this idea. Let we develop virtualisation layer > for Mac OS X versions of applications (because they are faster) > and then do such kind of 'porting'. > > This way we would have faster execution (Mac's applications > are more optimized), we do not need guest OS (avoid legal issues), > and we would have lower overhead. If we succeed to develop such layer > which adds max. 10% overhead comparing when same applications runs on > its native OS (Mac) without any virtualisation, then it could attract > more people to use OpenSolaris. A few things: 1) Many people have kicked around similar ideas over the years (maybe not on this list, though). It's a technical headache and requires lots of reverse engineering. Apple's docs are nowhere near the quality of Microsoft's. And even with lots of great talent, many users, and generally good docs, Wine is nowhere near "finished" after all these years. Something targeting OS X would be fraught with peril, especially since Apple doesn't care about back-compatibility like Microsoft does. Even worse if it's designed to only target Solaris. 2) The applications are not more optimized. :) There are some technical benefits to the operating system which you would lose, and would likely see a huge performance hit. 3) There are still quite a few apps which are not "Universal" and run PowerPC code via Rosetta. Emulating what a Mac does on the PPC platform is something that a few brilliant people have been working on for years without much to show for it. 4) Frameworks... This is pretty much a Mac-specific concept, and "properly" handling them will likely chew away many hours of a potential developer's time and sanity. 5) While we're on the subject, whatever happened to WABI and PWI? ;) (Rhetorical question to point out that Sun may have desires to avoid this area, due to past experiences. Please don't take it as flame bait.) > Once when we started to increase our user base significantly, then > companies who produce application software will start to port > their applications to OpenSolaris and we will have OpenSolaris > native applications with zero overhead. This will, in turn, attract > more users, which will in turn, attract more application vendors > to port their software portfolio to OpenSolaris. This is actually not really how it works. :( If you lower the barrier to entry low enough, you'll never get "proper" ports of software. Changing a company's development target to some API/ABI compatibility layer doesn't really gain OpenSolaris anything unique. Especially if it's something like Wine where other platforms (such as Linux) have better support (in the kernel, at that) for it. > This way, we could do transition from negative feedback to positive > feedback and take market segment which currently belongs more than > 90% to Microsoft. In my opinion, Microsoft is not the proper target right now. Right now, the Linux market is the proper target. More specifically, the "potential" Linux market. My feeling is that if you get OpenSolaris onto low-cost, decent performance machines with a good user experience, there will be adoption. And adoption increases corporate visibility, especially if it's something new users are adopting. :) Also, by targeting new users, there won't be any "unlearning" of "Linux does it this way, though!", which is way too common. (And yes, I know that Linux often does it "wrong"... ;) ) > That is all what I wanted to say. I hope it'll help to OpenSolaris. I'm not trying to discount anything you said. But with relatively limited community involvement at this stage, plus some technical feasibility issues, this isn't the stage to look at grand sweeping improvements which will probably not have any payout for 5 or more years. (That number was pulled out of a hat, yes, but ARDI was working on Executor for longer and never even got full Macintosh System 6 support, let alone System 7. While the operating system is arguably easier to target now, there's a lot more complexity than there was back then.) Hopefully our pair of posts will spur a healthy discussion on this list. :) And, as you probably guessed, I did lots of thinking, as you had suggested, while composing this email. :) Warmest, --Matt -- Matt Lewandowsky Greenviolet http://greenviolet.net/
