Michael Barr recently [September, 2010] wrote a column on The Sad State of Embedded Software Process, which tells us that the first embedded system was done about forty years ago, and we still are doing a collectively poor job at building them.
Perhaps there is a connection to using a forty year old language,with not having any great improvements in reliability after forty years?
Why can't any modern language gain significant traction in the Embedded Space? Whom has the clout to develop and promote a new, safe, language for limited resource hard real time (dead line can not be missed without catastrophic consequences) Embedded Systems that will actually gain traction?
I do not have those answers. I'm sure I could come up with a language, as I'm sure you could too, but how to get it used is the question?
Some of the 'safest' languages don't even make VDC's list, unless they are perhaps grouped into the 'Other' category of 11.5%, languages like Esterel and Erlang for example.
To me Esterel seems to be the most obscure language that is used in the most safety critical applications such as flight controls and railroad safety. Esterel has the company Esterel Technologies behind it, as well as an Open Source implementation of an Esterel Compiler. To get a feel of what it is like at the top end of the Safety Critical scale check out some of the Webinars such as Software and Hardware Foundations for Safety Critical.
Next in my obscure list of safe languages would be Erlang. Erlang runs a large percentage of the worlds telephone network, and is considered a 'soft' real time system (dead line can be missed without catastrophic consequences). There has been no larger and longer running networked system than the telephone company. I'm always amused to see such things as 'new' networking technologies rediscovering and reinventing problems and solutions that Erlang solved over a decade ago.
Because Erlang almost has what I'm looking for such as being small (in the sense of running on small micros), and meeting hard real time deadlines, Erlang is the one Functional Language that I play with the most, but does not meet those two criteria.
- Functions have no side effects; a function can not change a global variable for exmple.
- Data is immutable. Variables are single assignment; Variables don't change value after being set the first time.
Eliminating side effects can make it much easier to understand and predict the behavior of a program, because you are never asking yourself questions like "What is the value of X right now?", X can only be assigned once, so you always know its value.
Professor Stephen A. Edwards whom is a tenured associate professor in the Computer Science Department of Columbia University, once told me that freshman students to his class that had no programming experience picked up Functional Languages, and languages like Esterel quickly, while people that had experience with programming languages such as 'C', struggled to make the transition. Variables that never vary their value can be tough to get your mind around at first, if you programmed in 'C'.
Functional Languages where destined to live in obscurity until this week [Nov. 5th, 2010] when Don Syme of Microsoft announced that the F# Compiler plus Library Source Code is now available as a free download.
While F#,which has its roots in Objective CAML, has little hope of running on a microprocessor based Embedded System, Functional Languages just got a big push in to being more main stream with potential integration to things like Visual Studio. It is my hope that soon there will be a micro friendly Functional Language, that gains widespread adoption.
Will Functional Languages save us? We can only hope.
In closing it is worth mentioning that UML comes in at 5.8% on the VDC report, so I thought I'd mention Miro Samek's new graphical modeling tool QM, that I'm just starting to learn myself. The QM tool is available now for a free download and free, unrestricted use.