Last week I signed up with TekSavvy, intially with the intent of using the same network I was on, but without the burden of userous fees and stupid caps, but it turns out that Rogers is not cooperating with the resellers in my community, and TekSavvy called to say that in spite of the CRTC’s ruling to the contrary, they cannot resell the cable service owing to Rogers lack of cooperation. No problem, I simply returned the DOCIS 3.0 cable modem I purchased from Canada Computers (thanks guys for being so helpful), and I signed up for the 25MB/7MB DSL service from TekSavvy.

Just for fun, I ran a couple of speed tests on Rogers, and found their high speed service topping out at about 700Kbs download, and upload just a bit faster. So much for the “hi-speed” in the Hi-Speed product. TekSavvy is installing on Tuesday, and I just can’t wait. Rogers good bye, and good riddance — YOU TRULY SUCK and I won’t be back!

One thing which bears specific comment, are the efforts to support the Open Firmware device tree both within the U-Boot firmware, and in the Linux operating system. The device tree describes the SOC “bus” structure of the on-board devices, and possible the ancillary devices implemented upon a given board. In theory, this is a operating system agnostic description of the hardware which can be used by generic device drivers written to support the “flattened device tree” (FDT) model.

I haven’t fully realized all of the benefits of the device tree model yet, but as I consider the implications, I begin to fully appreciate the value of this approach. I suspect that this is a topic which will grow in importance not only in my own future work, but with any luck, in the future of all embedded development. Be sure that I will speak further on the device tree in future, and in the meanwhile, I hope that now that it is on your own radar, that you will give device trees the attention that they deserve.

Some time ago, I wrote my first C based Forth, which while anathema in some circles, afforded me the opportunity to play with forth in a hosted environment. More recently, I re-wrote the Forth, still in ANS-C, but designed it with the intention of making it very portable, variable word size, and most importantly, run natively on hardware which is sufficiently capable (Arm, AVR32, and even the more sophisiticated 8 bit SoC’s).

Forth purists would decry the fact that I implemented what is often referred to as a Fig or classic model, and more that I didn’t implement it in assembler, or even better, a Forth based umbilical system, but I did manage to accomplish my goals of a lightweight, extensible scripting language which could be used either hosted (Windows, Linux, BSD, and MAC/OSX), on 16, 32 or 64 bit, ARM, PPC, or Intel in its basic form without modification.

This miniforth is currently a glorified toy, but it I hope that it has the potential to evolve into a useful scripting tool in its own right. For now, it is certainly useful as an example of a portable C program, a primitive Forth scripting tool, a sandbox to learn Forth or an embedded scripting language, though that will take a bit more work.

If you’d like to play with MiniForth, simply download the single source file listed in the Linkroll on the right, or click here, and compile it with your favourite ANS-C compiler.


gcc -o mf MiniForth.c
clang -o mf MiniForth.c
mingw32-gcc -o mf MiniForth.c


Comments can be posted here, and if not spam, then I’ll publish them. As this language evolves, I’ll release under a permissive license (non-GPL), and hopefully, it might help put the Forth language on a few more radar screens.

Co-incidentally, Nova ran a documentary this week on Watson, and Dave Ferrucci’s team at IBM. Nova’s backgrounder was even more impressive than simply watching the performance of Watson’s avatar on the game show itself. Obviously, Watson’s talent was not easily come by, and only those who don’t fully understand the difficulty of the problem would have attempted it. Cue Dave Ferucci. Sure, critics might scoff, and say that the machine is not fool-proof, but by any measure, Watson is performing in a manner consistent with the best champions in Jeopardy history. No mean feat, even if the computer is the equivalent of over 6,000 powerful modern personal PC’s.

I’m not going to wax philosophic, I’m sure many bloggers will be chewing this one to cud, but I will touch upon one simple and blatantly obvious application of such powerful computing, and that is the field of medical diagnostics. Feeding the symptoms of known disease courses, and developing rules by using the talents of the experts in the field, the classic idea of an expert system might actually be achievable. What this will ultimately accomplish, is to bring the world’s best medical opinion to the masses. This may not happen in the immediate future, but obviously, Watson has legitimized the area of machine learning, and applied expert systems as not only feasible, but readily demonstrable. Kudos all around to IBM, Dave Ferrucci’s team, the Jeopardy people, and Nova for making us aware of this important and significant event.

What’s my point? Well, for years I have been using gcc as my de facto lingua franca, but yesterday, I installed CLANG from the FreeBSD ports, and compiled my bu-D (for backup to [local] disk — pronounced ‘Beauty’) package with CLANG. Immediately, several warnings pertaining to recent standards came to light, and in a matter of about 5 minutes, I repaired the BlobStor library, and the bu-D top end. Most of the problems were simply style related warnings, but in repairing them, I felt that the resulting code was that much cleaner, and more robust. So once again, my point? I like the diagnostics and the speed of compilation of the CLANG compiler. Full stop.

I’ll be cleaning up the Q-Schema library with the aid of CLANG over the next couple of days, and I have the strong impression that the end result will be a truly “carrier grade” implementation of my backup tool. This is increasingly important as the alpha test program for the CAStor cloud project approaches. More on this project to come, but suffice it to say that integration tests are going well, as they must be to see me futzing about with a new compiler, and currently four distinct platforms are supported, with Windows and Mac OSX to come.

I offer services to clients where I can recover data from sectors gone bad using a software approach which can re-calibrate the heads, and uses a persistent read to capture all but the most stubborn disk failures, but a head crash is a whole other kettle of fish. In this case, clean-room techniques are used to transfer the platters to a similar drive mechanism, and then intensive data recovery techniques are applied. I don’t have the facilities for this type of recovery, and my heart went out to the poor fellow, because the drive contained the sole copy of his family photos going back some years.

As I’ve indicated in previous posts, hard drives being mechanical devices, will fail at some point. All of them. The key to data recovery is not just backing up to a removable device, but keeping redundant copies. In this case, an educated consumer figured he was doing the right thing by moving his valuable data onto a backup device, but by removing the original photos from the computer, when he lost the backup drive, all was lost. Stay tuned, as I ‘m working on a backup system which will resolve most of the problems with backup systems in general, and will offer the kind of solution which will allow you to sleep peacefully.

No problem, I figured that I’d just move the relatively new hard drive (a previous favour when I used g4u to clone his dying hard drive onto a new larger one, and better still, I left the remainder of the disk unallocated). I figured (correctly) that XP Home would blow its cookies when it woke up on different hardware, but I figured I’d simply do a “repair install” over top, only after I backed up the hard drive of course.

The problem then became trying to find an XP Home install CD from which to install, and all I could find was an old XP Professional CD. Ok, I’ll do a straight install onto the unallocated remainder of the hard drive, and use the Microsoft utility to move “users and settings”. I had to re-install SP3, and some other updates as well, but that wasn’t a problem. On the surface, all went well, but after delivering the brand new and improved, faster and smaller computer, it seemed that the Outlook Express address book didn’t make the trip. Pardon me? Users and Settings apparently doesn’t include the Outlook address book! Go figure.

Apparently simply moving the .WAB (address book) file is not enough, and there are some registry settings which force the user specific address book to be recognized. I’d initially just tried to export the old address book, and import it over top of the new one, but I chose tab delimited file instead of .CSV (since XP Home didn’t offer to export as .WAB), and Outlook choked. Moving the .WAB directly had Outlook agree to import it, and then left the address book unchanged, even though it claimed to have successfully imported the 174K of data — but lied. Worse than any of that, all of the file bindings for Acrobat, Powerpoint and various other extensions were left behind along with the applications Ok, fine, there are hundreds of web pages which describe fixes for each and every one of the problems I encountered, and people who make a very good living out of Microsoft’s shortcomings, and given the market for broken computer services which Microsoft single handedly created why should’nt they?

Good software is generally written in a consistent manner with a set of reasonable defaults and consistent handling of configuration and settings. Good operating systems allow one to migrate licensed software from an older machine when the hardware is refreshed. Good commercial software doesn’t let the license manager get in the way of the user. Good system software doesn’t rely upon third party products for its security because corporate types thought the internet was just a fad. Yes, its true, Microsoft simply appropriated the BSD 4.4 TCP/IP Protocol stack when they found themselves late to the internet party.

Good software does not have arcane instructions, hidden functionality and impossibly unintuitive migration strategies. When Microsoft refers to Windows technology as innovation (and to be fair to Microsoft I’m not even going to mention the Vista debacle) it is instead inconsistent, arcane and unintuitive, and that’s just the user interface. The poor Windows System Administrators will be forever in a place which can use more air conditioning and Windows Programmers have a whole ‘nother ring of Hell to suffer.

The worst part of this rant is that it all could have been so different. The history I know has Dave Cutler coming from Digital to Microsoft to design the NT operating system. By all accounts he did it right, including an hardware abstraction layer (HAL) which allowed easy porting of the OS to multiple architectures. Then the bad part — Microsoft (presumably Gates, or Ballmer or some other bright wit) insisted upon keeping the Win16 API (updated to 32bits as Win32), and so the look and feel, and the fifty million or so poorly factored API’s were grafted onto an otherwise fine piece of work, and the crappy bloatware which had so ravaged the Dos/Windows environment was perpetuated onto what could have been a work of art — but was not. And don’t get me started on THE DAMN REGISTRY, which I’m pretty sure did not come from Mr. Cutler!

Propagating Windows onto NT was almost as bad a decision as IBM choosing Intel’s 8088/8086 segmented architecture over Motorola’s MC68000 (with 24bit linear addressing) for the original IBM PC. Sure, they saved some pennies on each machine, but created a segmented memory model conundrum for software developers for years to come. I suppose that’s what you get when marketing gets to tell engineering how to do their job.

If after reading this you believe that I have some antipathy towards Microsoft, you’d be right. It’s not that I hate them, I don’t. What I really hate are the incredibly stupid masses of PC users who actually believe that Microsoft has set a de facto standard simply because they have the market share, and who keep buying inferior products. If people started using alternatives in droves (and there are a bunch), then Microsoft would be forced to re-visit their crappy software, and fix it! Alas, the market leader has made DUMMIES of us all!

At full eclipse, the moon appears a mars-like red colour, and as I varied the ISO settings, the exposure time varied from about 4 seconds to almost 20 seconds. This was one of the reasons for the blurred images, perched as I was upon my back deck, which while braced with steel, did have some movement. Oh, well, on December 21st, 2094 I’ll be better prepared for the next time that a lunar eclipse coincides with the hibernal equinox 8-)!

Profusion has just announced that they’ve embedded E17 into, of all things, a refrigerator. Canonical, makers of Ubuntu have looked at E17 for its Arm platform, and there are already some other commercial announcements where either E17 or the underlying Enlightenment Foundation Libraries (and therein lies the magic and for me, the appeal) have been embedded.

The EFL libraries have been carefully handcrafted, and indeed re-written over the past several years, and hence the delay between the E16 and E17 releases. Starting with libimlib’s replacement with Imlib2, the developers of E17 have carefully developed a layered foundation, the key to which is the Evas canvas manager (libEvas). By itself, Evas does nothing, but the simple C API allows objects to be described to the Evas canvas manager, and then manipulated by simply changing its state. This has lead to efficiencies using libGL where possible, and in the case of Arm, where the hardware acceleration may be absent, though even so, rendering performance is more than acceptable.

Adding to Evas are Edjie, Ecore and Eina which make up the E17 stack, and added the widgets, theme-ing meta-objects and user friendly icing which will make E17 into a fast, flexible and embeddable window manager which will be on tablets, OSX, Linux, BSD and apparently, even household appliances.

While the big guys (Google) are using unwieldy Java frameworks glued onto phones and tablets (Android), and HP pushes a javascript development environment for PalmOS, and MeeGo uses a Qt based development model, I truly believe that whoever uses the E17 EFL platform as the basis for their embedded products will have some serious performance advantages, not to mention some really cool eye-candy for their bright shiny object and more important than any of that,  gain  a real edge over the competition.