Soft Watch

Electronics Times, April 30, 2001

At the latest Embedded Systems Conference in San Francisco, the Embedded Linux Consortium (ELC) decided to try to put together a specification for a standard embedded Linux.

There is a sense of deja vu about the attempts to create a standard embedded Linux. Anyone remember the mid-1990s and the brouhaha over the embedded Posix standard? More recently, we had the EL/IX proposal from Cygnus Solutions, shortly before being bought by Red Hat Software.

EL/IX borrowed much from the Posix activity and created four profiles, each successive profile being larger and more `complete'. There was nothing particularly wrong with the approach but it seems to have failed to have captured much interest.

If there was a widespread demand for a standard embedded OS, we would have one today. We have some with a good market share, but no dominant standard. There is a good reason for that.

There are very few embedded applications that are driven by standards. Those that are work to the familiar PC business model, where you need a wide variety of ready-to-run applications software. Even then, companies such as Microsoft and Sun Microsystems have defined profiles to fit a given market.

Companies in the ELC have to ask themselves exactly why Linux got a foothold in the embedded space. And maybe ask the engineers who implemented those systems. They would find out that it was because, given the source code, they could alter the OS and add things they needed.

It's not a strategy that suits everyone. Which is why the embedded OS market splits fairly neatly into halves.

There are those who buy binaries from their favoured supplier and work to that `standard' programming interface. And there are those who prefer to roll their own. Some of those people are pulling together a kit of parts from open-source software. That means not just Linux but RTEMS and the original version of uC/OS.

It is easy to see why the members of the ELC would like to see a standard. They fear the situation that developed with Unix in the desktop space. But it is driven from the perspective of the OS and middleware vendors who would like to create a wider market for support software.

There is a space for a standard Linux binary sold as a product, as much as there is a space for QNX, VxWorks and Windows CE. But such a product does not encompass all the applications for Linux and open source.

As a result, an embedded Linux standard or, more likely, a set of Posix- derived profiles, can gain some traction but probably only as much as the Posix embedded profiles did.

If the community can sort out some issues between Linux and Posix, the user community may choose to adopt those interfaces, but as with anything in the open-source space, they will vote with their feet not in a consortium.

Copyright: United Business Media International Ltd.