Analyst: embedded Linux costs more

Posted Jul 16, 2003 13:15 UTC (Wed) by TimCunningham (guest, #10316) [Link]

http://www.embeddedforecast.com/WhitePapers.htm

Nothing says professional like "Untitled Document"!

Posted Jul 16, 2003 14:02 UTC (Wed) by ssharkey (subscriber, #4451) [Link]

Of course, they probably forgot to ask this question:

How many embedded projects have you done with this platform before?

And probably this one as well:

How familiar with the environment were you before this project?

So, assuming that most of their respondents were new to Linux and familiar with Windows already, the results could hardly be called surprising.

A more interesting question/survey would be to check back after the team is familiar with the tools and environments, and see if the results hold.

-Scott

Posted Jul 16, 2003 14:53 UTC (Wed) by pj (subscriber, #4506) [Link]

Hrm, I wonder if they factored in the per-unit Windows licenses in that?

Posted Jul 16, 2003 15:07 UTC (Wed) by betanio (guest, #12972) [Link]

http://www.embeddedforecast.com/EMFTCD2003v3.pdf

To download this "great and honest" document, it is necessary to create an account and to place the PDF on the basket.
The URL above is sent by email.
Perhaps this work is unnecessary: I do not agree to the estimates.

Posted Jul 16, 2003 16:05 UTC (Wed) by davidl (guest, #12156) [Link]

What do they mean by Linux? These is not defined. If you develop 'embedded' applications with Linux, from scratch, then yes it will take longer. Using Montavista Linux, and development tools like Embedded Qt it will take no time at all. What about J2ME on Linux? This is being used everywhere in mobile phones for instance. What tools are being used with Linux? The only thing mentioned was VS.NET, and I know that doesn't run with Linux. The only thing it mentions with regards to Linux are 50 anonymous OEMs (not developers or companies - OEMs - a Microsoft term) using various 'distributions' of 'embedded Linux'. Hardly a definition. Anyway, I doubt whether anywhere near 50 OEMs are using Microsoft embedded tools in projects of the size and cost they seem to be talking about.

Notice how the figures of time-to-market, the average number of developers and the cost(!) are simply increased without any explanation. "Yer, we'll add a million to it. That sounds good". Of course, no Microsoft-sponsored report would be complete without the "We're not slinging mud at Linux, honest!" line:

"Embedded Linux can be effective in distinct scenarios, but does not broadly translate into lower cost or faster development,"

I couldn't get access to the site, www.embedded-forecast.com - which bizarrely, gets forwarded to www.embeddedforcast.com - as it was down. The latter is a shiny new Windows 2003 server running IIS 6.0, so no shocks there. They should have stuck to www.embedded-forecast.com which apparently is running hosting software on Solaris 8. Something can be inferred from that, I'm sure, such as a feeble attempt to mask what they're using or a 'Microsoft takeover'. Anyone's guess. Wow, it's now back up and it's got the look and feel of a Microsoft site. When I try and download the PDF report I am presented with a certificate warning which tells me that this certificate is registered to registeredsite.com not embeddedforecast.com. I accept and, yes, the ubiquitous IIS 404 error which tells me the document cannot be found. These people really know what they're doing. The site is now down again. Hilariously outstanding.

This report (from what I've managed to read) is meaningless as it doesn't define what it is talking about. It just talks about 'Linux' as a one-size-fits-all thing, like Windows, which it isn't. This is exactly the sort of concept that Microsoft is having no luck whatsoever in getting it's head around.

It is yet another transparent Microsoft-sponsored report that will simply be forgotten. It's simply another desperate attempt to get Microsoft software accepted in the embedded (small devices) market, and to turn it into a PC market clone where companies simply make devices and pay an OEM license to run Windows. Notice how Pocket PCs have to be more powerful and more expensive than anything else just to get it to run Windows. It hardly qualifies as an embedded device in my book.

This isn't happening and it isn't going to happen because companies like Nokia and Sharp aren't thick, they have good dominant products already and they want control over the development of them. Linux, Symbian and J2ME in particular are entrenched in the embedded development market and they are not budging. I'm wondering when Microsoft are going to learn that they will not make a monopoly out of this market. They are still kicking and screaming in that stairwell I'm pleased to say.

Posted Jul 16, 2003 16:43 UTC (Wed) by kokopelli (guest, #11341) [Link]

I'm not a trained seal and I don't jump through hoops to get untitled documents, but I wonder how many subtle biases were in that study.

The most obvious has already been addressed - are they comparing experienced Windows developers working in Windows vs. Linux, or experienced people working in their own comfort zone?

But I've worked on enough embedded projects to have an even more compelling question: what are they trying to do? Could the Windows projects have a smaller scope? Could they be creating embedded devices that have to interoperate with specific proprietary systems? Could the project itself be selected with an eye towards the tools available for that platform?

Finally, if this is true then why is the marketplace so clearly going the other way? It looks like stovepipe systems are finally dying, but the market seems to be going towards Wind River and Linux-based systems (e.g., Lynx became Lynix or something like that), not embedded Windows. It's not like Microsoft has no name recognition in this market - one of most annoying things about embedded development is that so many of the cross-compiler tools are needlessly platform-specific. (At my last job we actually used the Gnu toolchain for the cross-compiler, but we had to use Windows because they only ported a few of the cross-compiler tools to Windows. "Who does development on Unix/Linux systems?!")

Posted Jul 16, 2003 21:39 UTC (Wed) by kfox (guest, #4767) [Link]

It seems the author found that Linux
projects are bigger than Windows projects.
Companies work on them longer and assign
more engineers to them.

The author also found that engineers
working on Linux projects earn a higher
salary.

From this he divined that Windows is
the cheaper platform.

Here's a different hypothesis: Companies
are more likely to choose Linux for large,
ambitious projects -- just the kind of
projects that need the best people.

- Ken

P.S. The dollar figures for maintenance
and support are so tremendously skewed
in Microsoft's favor that IMHO either
the author has misrepresented his numbers,
or Microsoft is yet again using monopoly
profits to dominate other markets.

Posted Jul 16, 2003 21:47 UTC (Wed) by emk (subscriber, #1128) [Link]

The embedded marketplace is incredibly diverse--dozens of processors, operating systems and compiler vendors, with thousands of different niche markets. It's not easy to survey meaningfully.

There's a real danger in confusing "embedded" point-of-sale applications with mobile applications with medical devices with industrial controllers with exotic niche applications with whatever else is floating around. Most market "surveys" lump together a hundred dissimilar markets and produce meaningless conclusions.

For example, Linux could easily be more expensive because it's being used in more technically aggressive projects requiring more custom development than is typical for WinCE. Or it could be that WinCE is preferred for certain point-of-sale applications, and Linux is winding up in more mobile devices.

Really, it's like trying to compare Ford, Boeing, skateboards and NASA as methods of transportation. You have to be a bit suspicious of any analyst who even tries, especially if they produce obviously partisan results.

Posted Jul 17, 2003 13:45 UTC (Thu) by ccchips (guest, #3222) [Link]

All you have to do is remember the Heinz ketchup commercials from the 1960's to understand where these so-called "analysts" get their data for comparison.

For awhile there, back when very large numbers of Americans were getting really tired of false advertising, the Federal government actually had people on staff that *tested* the advertisers' claims. It didn't take the private media operators long to start discrediting these watchdog organizations, event othe point of claming that the workers never showed up for work (of course, they would pick some really glaring example, and the rest of the hard workers could go pound sand.)

Maybe we need to start privately funding an organization that will watch the other privately-funded organizations and make sure they tell the truth.

Note: One of my friends in college told me that he was in a marketing class where the teacher talked about the ketchup commercial. Apparently, Heinz was putting their competitor's product in the oven and their own in the refrigerator, and then demonstrating the thickness by dumping the bottles out at the same time.