Slashdot Mirror

← Back to Stories (view on slashdot.org)

ISI, Mitsubishi to Develop New Operating System

Posted by Roblimo on from the anything's-better-than-wince dept.

Richard Finney writes "The Associated Press is reporting that Integrated Systems Inc. and Mitsubishi are teaming to create a new operating system in a Yahoo! News story here. The OS will be for new, portable gadgets that will feature high-speed Internet access."

13 of 67 comments (clear)

  1. Java was made for PDAs by ssorc · · Score: 2

    Here goes.

    There have been a lot of comments slating this new OS, mostly along the lines of 'Linux Rules'. While I run Linux on my PC, I'd never run it on a PDA.

    Firstly, I don't need a multiuser operating system for my PDA. Why waste precious space with useless code? The same goes for other great Linux features.

    Secondly, James Gosling wrote Java for Cosumer Electronics (CEs). Sun was looking for a new direction. So they got a bunch of people together and decide to make a new operating system for CEs. Gosling then invented Oak. Sun decided that Oak made an awesome programming language, so they renamed it Java and the rest is history. (Ok, so I simplified a little). Some of the old CE vision for Oak still shows in Java, making it idea for this kind of programming.

    Thirdly, if you have a processor which runs ByteCode (i.e. runs compiled Java natively) you can write an GUI plus OS which takes up less than 1mb. Sun did this a few years back, I think. I know there are some pretty small Linux installs but I don't know of any that small.

    Caio.
    Simon.
    [Web page currently down due to selling of server]

    --
    /-\-/
  2. Cygnus' eCos by jbgreer · · Score: 2
    I'm amazed that prior posters haven't mentioned eCos ( http://www.cygnus.com/ecos/ ) as an alternative to this. To quote their page:

    Cygnus developed eCos to provide an open, standard infrastructure for embedded developers worldwide. eCos provides a common run-time software environment that can be configured from source code to meet most application-specific requirements. eCos is available as an open source product.

    Cygnus delivers eCos from sourceware.cygnus.com as an open source release. Cygnus also delivers eCos as a commercial product backed by Cygnus developer support.

    p.s. Just to make it clear, I have no affiliation with Cygnus.

    --
    The Norton Anthology of English Literature, 4th Ed., Vol 2
  3. About ISI... by redwraith · · Score: 2

    ISI is the vendor we purchase our development system from, for use in a real-time audio processing system. The OS is pSOS, and bundled with compilers, tools for emulators, and performance processing tools. While speaking with other departments who use the development system for larger efforts, it was pretty clear to me that the tools and add-on developments (particularly profiling and code navigation tools) were feature rich and EXTREMELY bug-ridden, to the point were there is almost no useability. Feature-creep and glossy handout material.

    pSOS seems a fairly capable OS, but we have hardly put it through its paces. I hope that if ISI is indeed attempting to come out with another embedded OS (and they have purchased the company that developed pSOS), that it will be that end of the company doing the development, not the hacks that are cranking out those sad tools.

    ISI (IMHO) is basically a system integration vendor trying to move on to development, and doing it with a marketing-driven twist, instead of actually shipping code that works (shades of Micros~1?). I don't expect much to come of this, if they can't produce tools that work, they won't compete in a field that appears to be saturating.

    Feel free to take this with a grain of salt, I've spent quite a few hours being jerked around by an ISI salesman trying to sell me tools I don't need that apparently don't work.

  4. No, No, We VERY MUCH NEED New OS Designs by Christopher+B.+Brown · · Score: 3
    If you look back at the last ten years, there has been the anomaly of there being only a small amount of real progress in the development of operating systems.

    Twenty years ago, there were a goodly number of derivatives of Multics as well as attempts at "truly new" stuff like Hydra on their way.

    In the '90s, the only alternatives anyone has had any "faith" in have been Windows NT, and, in the last year, Linux. I would contend that this has been the result of

    • A concerted attempt by MSFT to buy out the OS research groups,
    • The failure of IBM's "WorkPlace OS" project, which is related insofar as it was based on Mach, and MSFT offered Rashid, the Mach architect, "too much to pass up," and
    • Related to both of the above, a conscious attempt by MSFT to convince the marketplace that the future would involve "Windows Everywhere."
    • Add in the factor that there used to be sizable projects that combined academic, government, and corporate funds to produce public goods, as resulted in such things as X, BSD 386, CMU Lisp, and Andrew, that don't seem to be continuing to be sponsored
    And you get the situation where it looked like the world might conceivably be "Windows Anywhere."

    The way we get improvements, in the long run, is by trying a diverse set of different things. Linux is not appropriate everywhere, and we may find, out of some alternative OS research, the successor that will be so much better that it can ultimately replace Linux for "general purpose" applications.

    Furthermore, there are more kinds of applications out there than merely those for which the "big server" that Linux is ideal for.

    • For high security applications, the models provided by the UNIX security model are starting to show creakiness.

      Experimentation with capability-based systems is needed to figure out how to build more secure systems.

    • For very small embedded systems, Linux is simply too big.

      Cygnus' ECOS has a kernel that can be configured as small as a couple of kilobytes.

      It may be amazing enough that you can boot Linux on a PalmPilot; that doesn't mean that it is actually useful to do so.

      If I wanted to build something to fit in 2MB of RAM and some smaller amount of ROM, I'd not pick Linux.

    The world will be served by allowing Linux to be replaced by a set of OSes that are better, once such systems can mature into usefulness.

    Remember also that Linux may be legitimately regarded as being merely a kernel.

    --
    If you're not part of the solution, you're part of the precipitate.
    1. Re:No, No, We VERY MUCH NEED New OS Designs by Tet · · Score: 2
      Experimentation with capability-based systems is needed to figure out how to build more secure systems.

      Agreed, but much of this work has already been done. See the DG/UX B2 security option, for example. Other proprietary Unix vendors have done similar things (e.g. Trusted Solaris), but AFAIK none of them took it as far as DG. Interesting to note that DG/UX is one of the few innovative Unices out there. Although originally an SVR[34] licensee, later versions rewrote the kernel from scratch, in a similar manner to Linux, to support the additions they needed. Sadly, I think EMC's recent purchase of DG will result in the demise of DG/UX.

      I definitely agree that the world needs new OS designs. The more, the better, as far as I'm concerned. Nothing promotes innovate as much as competition.

      --
      "The invisible and the non-existent look very much alike." -- Delos B. McKown
    2. Re:No, No, We VERY MUCH NEED New OS Designs by Brian+Knotts · · Score: 2
      I obviously agree that:
      • Linux is not the solution to every computing problem
      • Research into new operating systems needs to continue, and perhaps be accelerated
      Looking at some of the comments from developers working on embedded systems, it sounds like we can, however, take one thing from the "Linux experience:" we perhaps don't need a whole lot of new proprietary operating systems, tiny or not.

      You even used the example of eCos, an open source RTOS for embedded systems.

      It's certainly an arguable point, as many, many people still support the proprietary software model. Even open source advocates support it for certain applications. But to suggest that open source is a model that should be followed, in the interest of consumer and developer alike, is at least a legitimate view, whether one agrees with it or not.

      BTW, I found your comments about the Workplace OS of great interest, as I had not heard about that particular event (Microsoft hiring away an important developer) before. Do you know where I can read more about that?

      --
      Interested in XFMail? New XFMail home page

    3. Re:No, No, We VERY MUCH NEED New OS Designs by Brian+Knotts · · Score: 2
      Thank you for providing the links. I've only been able to scan through quickly, but there appears to be a great deal of excellent information, particularly at the first link. I'd suggest others take a look.

      Linus' views on microkernels are, of course, well known. I think he's right, to a point, and in the context of where computing is today. Projects like the Workplace OS, ambitious as they were, I think reflected the hopes of many of where the microkernel could take us (I certainly was all caught up in the hype for the project at the time). Perhaps they still will. I can only dream.

      --
      Interested in XFMail? New XFMail home page

  5. Bad reporting by AP; not a new OS at all by Zach+Frey · · Score: 2

    The AP news report cited claims that ISI and Mitsubishi "will develop a new operating system for hand-held computers". However, the press release from ISI claims that WebPDA is based on ISI's existing pSOS+ RTOS.

    A basic breakdown of what WebPDA actually is:

    • pSOS+ as the operating system
    • pJava from Sun as the JVM
    • Mitsubishi's VRPC hardware design
    • Espial's Java-based desktop and applications

    So, rather than a "new" OS, these companies have bundled together an existing RTOS, an existing JVM, an existing PDA hardware platform, and an existing Java framework, and made a big marketing announcement out of it. Big whoop. Every major RTOS vendor competing with ISI either already has a JVM on their platform or is working desperately to get one, and will have one Real Soon Now(tm).

    It is nice, though, to see the traditional RTOS vendors not quietly ceding the PDA space to the 800-pound gorilla of Microsoft and Windows CE.

    Was Thomas Alva Edison the first 20th-century entrepreneur, a man who contributed greatly to the shape and form of modern society, and was therefore the spiritual godfather of Bill Gates? Or was he another 19th-century corporate baron, a public figure whose greatest talent was self-promotion, and was therefore the spiritual godfather of Bill Gates?
    -- Brian Santo, EE Times, 7/14/97

  6. Re:Real-time OS needs to be on-chip by Burnon · · Score: 2

    I don't know if I understand this comment. A lot of ASIC vendors put ROM and/or RAM on the ASIC along with the microcontroller. Unless you're talking about some of the interesting new "threaded processor architectures", people have been storing RTOS's in on-chip memory for a long time now.

    The real question is, why do you want the OS on the microcontroller itself? If it costs you more to make a big ASIC with embedded memory to get the job done than it does to make a small ASIC with external memory, your company will go with the cheap external memory. Otherwise, your company will go with the on-chip memory.

    However you choose, there's nothing magical about putting the RTOS on the chip. If you've enough RAM/ROM, it works, otherwise it doesn't.

  7. Lordy, not another itty bitty OS by shambler+snack · · Score: 2

    Let's see, we have:

    Palm OS from 3COm.
    EPOC from Psion
    QNX from ONX (think of Amiga and Proton UI)
    OpenDos from Caldera Thin Clients a.k.a Lineo
    Embedix, and embedded Linux from Lineo
    TRON, a series of embedded OS' from Japan

    And that doesn't include Windows CE from Microsoft, and I'm sure I've not even scratched the surface. This sounds like another closed Japanese "innovation".


  8. More links by grungeKid · · Score: 3

    ISI Releases Net Device Reference Design
    Trio seeks to jump-start Java-based PDAs

    It looks like a cool system... and before you moan about it being java based and therefore intrinsically slow, remember that java was originally based for appliances like settop boxes and handhelds. They're partnering with Espial for java class libraries, and Espial has done some great java libraries with a very small footprint. Will be interesting to see if it will be interoperable with the Palm + KJava

  9. It's all about choice by howardjp · · Score: 2

    It is good to have choice. Wasn't that what Windows Refund Day was all about? Think about it, Linux (despite what you might think) is not perfect for everything. Nobody I know would give a Linux based system to their grandmother. I personally do not use Linux at all because it fills no need of mine that something else does not fill better. Cut them a break.

  10. PSOS and a longer OSS rant. by nyet · · Score: 2

    I agree whole heartedly with all your comments wrt PSOS. We've gone through the exact same headaches as you described. PSOS is a nice, solid, well engineered kernel. It is small, and much less complex (albeit less feature rich) than some of its competitors, like WindRiver's vxworks. What is BAD about it is their draconian license polic - per "instance" royalties, and a LARGE entry cost, not to mention the legally non-reverse-engineerable binary kernel itself. Interestingly enough, the compiler of choice for the i960 (our old platform) was Intel's ctools (their free hack of gcc). The rest of PSOS's targets all required payware (read: per seat, lmgrd-guarded licenses), as did their debugger.

    Also, the core was a complete black box (binary) with a whole TON of hooks into it so you could tweak it for your application.

    Tech support was a nightmare.. Getting even the most rudimentary technical information from them usually required 3 weeks and an NDA or two (like their task database structures in UN-documented .h files!)

    We have since decided to drop them for new projects and pretty much demand that vendors meet two requirements.

    1) no royalties
    2) ALL source code

    For those of you who doubt that the efforts of the FSF and those OpenSource wackos made any diferrence, you really have no clue.

    The state of embedded OS devel is a strange one. Most engineers that do this kind of design and coding are used to running their little debugger and compiler on their win95 machine. There is a growing segment of this market, however, that understands the big picture, and is willing to learn a bit about UNIX, mainly because of one thing: TCP/IP. This is remarkable, honest. No engineer in his right mind would be satisfied with what MS tells him networking is, givin the history of BSD sockets, and the fact that even the first winsock stack was not actually written by MS (seeing as they thought TCP/IP was a toy).

    Now that just about every embedded device can (and will eventually) be network aware, the feature space of these embedded OS's have much more in common with UNIX than they do their ROM-BIOS ancestors of lore. What does this mean? Now that most people's first UNIX experience is likely to involve Linux, they are also likely to be exposed to what OpenSource REALLY is - NOT corporations trying to get a free ride from random outside hobbiest/programmers, but a cooperative effort designed to reduce the amout of time duplicating somebody ELSE's efforts.

    How? By a)demanding source from the vendor (it may not be OSS, it may be subject to NDA, but that part is irrelevant so far) and b) actually cooperating with others doing similar work on similar platforms.

    Once the uninitiated engineer first discovers what its like to exchange source level bug fixes with other engineers, there is no going back, no matter how dense or MS centric they may be.

    One final comment. People on slashdot like to spout endlessly about "oo lets just put Linux on the toaster". They dont realize what it takes to get a totally foriegn, customised, embedded system running. You cant take ANYTHING for granted. You have no fixed memory map. You have no display, let alone debugged serial drivers. Your DRAM controller may be completely non functional. In this realm, products like PSOS and vxworks shine.

    Linux is PERFECT for embedded applications, but somebody has to do the booting. If the target is a brand new mboard from ASUS, with a known memory map, PCI controller, etc. AND there about 10,000 people all working at booting it and fixing bugs, great, linux is trivial.

    But if my target is some weird iron featuring the latest funky embedded microprossesor from motorola, there is zero chance of linux being ported to my target by somebody else, least of all motorola themselves. And the in-house R&D guys dont want to spend 4 months getting the OS booted, and then wondering how reliable the whole thing is in the end.

    IBM's release of the PowerPC reference design may change this, since it should be easily embeddable given the fact that both IBM and mot have some truly excellent embeeded PPC cpus coming out.

    BUT! I need the damn boot roms for the friggen evaluation boards, or its no sale, and I have to shell out the bucks to ISI or Windriver for their environment.