Slashdot Mirror
Red Hat And Lineo Respond To MS Embedded Linux FUD
jeffy124 writes: "Red Hat and Lineo, the major spearheads of Embedded Linux, have said that Microsoft's recent white paper comparing Linux and Windows XP embedded is full of inaccuracies, false facts, and overall distorts the value of Linux in general. Lineo has gone as far to say it flat out lies about Linux. ZDNet has Lineo's response, Red Hat comments, and a summary article." Updated by HeUnique: LynuxWorks has also wrote a response (only this one is a bit more detailed).
What makes you say that Embedded Linux is "suitable" for use in ATMs/casino machines? I totally agree that the likelihood of Embedded XP taking off is unlikely, however at the same time I doubt any of the hardcore engineering crowd (that is use to the rock solid reliability of PLCs) is going to be touching Lineo anytime soon (yeah I'm sure there are some case studies/design wins where some cash was passed around and some token implementations took place in non-critical sectors).
If I were to build a critical embedded system right now I wouldn't touch embedded Linux with a 40ft pole: It'd be QNX or one of the other proven systems that I'd trust. Let embedded Linux prove itself for a decade and then maybe then it'll be trusted, but as it sits it seems like yet a Microsoft like "try to get the same code to run everywhere" type of attempts.
I really hate this argument! At least with GPL code you have a choice; use the code and release as GPL or don't touch it. With Windows code, you have no choice; as you can't get the code without paying MS money or getting some NDA signed, you cannot use it.
Obviously, both sides are biased and think their product is better.
True, but Lineo didn't say "Linux doesn't support plug and play".
Not only is it blatently untrue, Linux DOES support plug and play, but it's supported it longer than NT. And XP is NT 5.1; if you don't believe me, check your web-server logs.
I have done porting work with just about every single embedded OS (some predating windows CE by about a decade). I have built/tested/deployed i960 and AMD29k boards using psos, m68k boards using vxworks, and ppc boards using nucleus, all in house, from scratch. You, sir, apparently have done NONE of those, because if you did, you would know that porting Linux is no different. We did it for three different ppc boards, also in house. The process was no different from developing a BSP for any of the other commercial vendor products. The difference is that Linux 1) is a real OS (even if it is lacking in the RT arena still) 2) no rediculous per unit royalty fees 3) free tools 4) portable tools (i.e. can cross compile from whatever platform I like) 5) rediculously stable 6) HUGE driver support (with source - something the commercial vendors love to nickle-dime you for, btw) 7) a large software library (e.g. just about every useful UNIX app exists in one form another) 8) excellent support via source code, email-lists, and irc. (Yes, the "linux has no support" meme is complete FUD - have you ever tried to negotiate WindRiver's support dept. even WITH a fully paid up support contract?)
The fact that MS has been "doing embedded versions" since the mid (try late) 90's is completely irrelevant. There are a ton of crappy, buggy, useless products that still exist today that their vendors have been "doing" for a lot longer. What makes you think that just because they are old that they don't suck?
There is no common integrated development environment (IDE) for Linux. OS development is command-line driven and applications development requires a new set of tools for each device. Developers must either build their own tool-chain from piecing together Open Source tools or opt for a specific vendor's costly toolset.
Uh, yeah, that's what happens when you use a monopoly to put all of your competition out of business. In the old days there used to be a choice of IDEs for Windows. In fact, I remember at one point Borland was the leader in development tools. Oh well, what do you expect from M$?
If you really read the white paper, it seems to me that the only real point that MS is making is that under linux you have a choice of what to include or use in your embedded device.
;) )
The windows solution they provide may be very good, but what if you don't want it? I mean, it sounds to me from the paper like if you use MS, then you are building an wireless internet wma/web browser/email device. They talk about how much better the web is on it and how much better the media playback is on it. Why don't they just release a device with all these features themself?
On the other hand, the linux solution they describe is not very concrete on what it can or cannot do. Seems to me like linux has options, while win XP doesn't. (except in the choice of languages aparently, you aren't contrained to java and C like in linux
If I were desgining an embedded device, I would use linux, cause even though it might be 'more work' according to this report, at least I'd have a choice of what tools and features to include in my device.
I am a software engineer writing networking code that runs in embedded environments. No one in their right mind would choose Windows XP for embedded development for a number of reasons.
1. Embedded devices use a wide variety of microprocessors, memory architectures, and hardware, often with custom boot code. At least in the hard-core networking industry, X86 is not a big player since RISC processors tend to be a better fit (almost all networking protocols are big endian, for example). As far as I know MS is X86 only (not counting Intel's Unobtainium).
Also, many embedded products are based around non-Intel processors to cut costs and power requirements. There are a wide variety of RISC processors out there with varying levels of integration.
In the networking area I'm in (dealing with network processors from companies like Agere, Broadcom, Intel, MMC, Vitesse, etc.) most of the parts come with support for Linux and VxWorks. I do not see much support for Windows.
2. Embedded developers often need very low level access to the operating system. I doubt very much that MS will make available the source code to their OS. Where I work, we have the full source code to VxWorks ($$$, but worth it), which our product is based. This allows us great flexibility in terms of adding features or fixing bugs (what, bugs in commercial software???). If we had to wait for Wind River (or Microsoft for that matter) to fix a bug we'd never get anything done.
3. Linux is becomming very popular in the embedded area in part due to its open source and the licensing issues in many instances. Linux also has a proven track record of having been ported to a wide variety of hardware, from IBM mainframes to the Sega Dreamcast and other platforms. XP has no such track record.
4. Many embedded environments do not want features like IE, media player, or even a GUI. In the networking products I've worked on the only front end is either through a serial port or a telnet session.
Embedded developers are not your typical programmers. It's one thing to write an app in a nice GUI front-end and be able to step through the code in a debugger (like one can do with user-space apps). In the embedded world it's often the case that everything is running at the kernel layer.
-Aaron
This post is encrypted twice with ROT-13. Documenting or attempting to crack this encryption is illegal.
Why Microsoft Windows XP Embedded and Not Wind River? -a lu ation/compare/notwindriver.asp
http://www.microsoft.com/Windows/embedded/xp/ev
It's exactly the same type of article they used to attack Linux. But that's a far less juicy story, isn't it?
But one has to wonder, is any of the stuff on slashdot FUD, opinionated and/or innacurate wrt Microsoft? (witness michael's article about MIME type holes recently)... mm... just a bit ;)
And, just to secure my Troll rating (and to prove that I too probably dunno what I'm on about), an OS which compiles into a 350K or so kernel for even the most basic of functions, plus tacked on realtime scheduling doesnt strike me as being very appropriate for embedded applications. Sorry folks, but this is one area where you NEED to pay for a lot of R&D and yea that does mean proprietary software; I'm no CS student but I do know that hard RT is a thankless thing to get right, as is supporting embedded microcontrollers and peripherals (CPU's dont exist in vacuum, right?) and consistent support for dozens of possible platforms (and, yes, bootloading said platforms). I remember RedHat was making something called EcOS.. it's young but the architecture at least seems designed for embedding; anyone know what's up with that recently?
Far be it from me to point at that you seem like a clueless, knee-jerk Linux zealot who loves to feel persecuted by Microsoft, but...
If you go to http://www.microsoft.com/windows/Embedded/xp/evalu ation/compare/notwindriver.asp, you'll see that they have the exact same type of article discussing Wind River. Gee, and it's even titled "Why Microsoft Windows XP Embedded and Not Wind River." Truly amazing. Sorry if I ruined your persecution complex. :)