Curt Schacker spins interesting thoughts about the embedded software industry on his blog. He claims that the industry is geared towards a service model, where vendors work against a Statement of work, rather than providing a product spec and an evaluable product upfront.
While I can see this happen in certain parts of the traditional embedded software industry, I think the offsetting trend will be the inflow of COTS (commercial off the shelf) products from the PC world, based on Windows, Java, VMware, and Linux, to name the most important platforms. These platforms commoditize the underlying hardware and eliminate its dependencies, as they have done in PCs, and thus enable economies of scale formerly unknown in the embedded software space.
The reason for my belief in this "convergence" is that working in a service model is very expensive and doesn't provide the economies of scale that a COTS product can provide. we're talking of a 100x - 1000x cost differential here. The additional computing power needed to support these later generation computing platforms will be mitigated by recent breakthroughs in computing power (Intel and HP just announced 100x breakthroughs, accelerating beyond Moore's Law) and storage (just look at NAND over the last few months).
As a result, a large part of the market traditionally being served by embedded systems software manufacturers will be covered by standard platforms known from PCs and servers and confine these traditional vendors to even more nichy spaces.
This convergence trend is also fueled by the fact that mobile devices (the largest embedded market next to automotive) become application platforms (while PCs start to become communication devices, just think of Web browsing and Skype).
Also, the connectivity between mobile/embedded software and traditional PCs and web services results in "spill-over" of technology from the more economical mass software markets to the cost-intense embedded systems space. In lockstep with that you'll also see more enterprise type developers looking into embedded software, because they "stick" to their technology. These new generation embedded software developers are often more sophisticated in leveraging COTS software than traditional engineers, who often origine from the hardware space and sometimes suffer a NIH syndrom (50% still write their own database - in the year 2007!).
db4objects is well positioned for this trend, serving a wide range of application with an "embedded" database (which is NOT a device database only, but one that's "invisible to the end user", in line with IDC's and VDC's definitions) -- from servers to PC to mobile devices. In fact only about 10% of our users are strictly speaking working on "device software" (like photocopiers or cars). Many more use db4o on PCs for device integration (e.g. the configuration console for an alarm system), on PCs or mobile devices to run packaged software, and even on servers, where the database engine gets embedded in a service oriented architecture.
So I am not contradicting Curt - he's right for the transformation of the existing industry. But at the same time the growth of this ever more specialized and cost-intense traditional approach will be much slower than the growth of PC-proven technology in what traditionally used to be a totally distinct market segment - that's the sweet spot targeted by a COTS vendor like db4objects to serve the embedded systems industry.
The laws of economics and Gordon Moore are on the side of COTS products.