Floid : QNX doesn't do the impossible with low mem. I forget the requirements for the demo disk, but they must've been at least 8MB, and it's no longer offered.
Hi Floid,
That's because it used QNX's 4.x-lineage products, which since early Phoenix days have been surpassed by the 6.x stuff. Incidentally, the reason this would even require 8 megabytes was for browser caches. I think the demo would run in considerably less though (4 meg I think, but it's been so long since I read about the QNX 4.25 demo and then tried it...)
There were people developing QNX-based products that used less - bring what ya need and leave the rest at home ; } ...There are and have been embedded products that ran in less of course, but these were based on some pretty limited OSes - some really had little that could be
called an OS.
From the download page, the current suggestions for what was the RtP (now "Momentics NC"?) are '400 MHz Pentium or better, 128M RAM , 1.0 G disk space.' I ran it on a Pentium 133, and managed to chop the system image (forgetting my terminology here... the nifty little filesystem-in-a-file that keeps your array of microkernel daemons or whatever as 'portable' as a monolithic kernel) down enough to survive in 32MB, but as soon as you want a desktop background, use of the Flash plugin, or a few pieces of useful software open... you're going to crave more memory.
There were more than a few people in Phoenix who got useable systems running in 16 meg (first place to check is buffers and cache defaults, then find the libraries you might never require, etc), some that got it down to running in 8 - and I think a couple of QNX vets got it running in 4 (still with the Photon MicroGUI, I believe!) ... So, even sophisticated products can definitely run on modest Flash - and that's with an OS that has
way more services than the old, simpler 4.x microkernel and process manager, etc.
You mention Momentics: As you surmise Momentics itself is the offspring of RtP, essentially being a developers' desktop complete with the GNU toolchain and a fair number of other facilities for development and personal use, has generous buffer defaults, etc. This makes it possible to comfortably design, self-hosted with QNX, for products that run in a lot tighter space - targetting multiple architecures with runtimes, natch - which I still think is the superior way to painlessly have top performance on many processor architectures.
(The "pro" version adds Eclipse IDE with extensive third-party tools, really deep systems analysis tools, custom libraries for embedded work, and lots of othjer goodies).
...Anyway, it's incredible what they've achieved in a few short years - but back to the memory requirements issue: as RtP (QNX6) alphas and beta progressed it became obvious that people were wanting more more more and that the price of memory and storage was becoming cheap enough to design cost-competitive products with greater features and facility. QSSL even pretty much shelved the non-MMU version, offering it only as custom work, since the median architecture for their OS was logically more sophisticated anyway.
Taking this back to AOS4 (or my fave before even QNX, MorphOS), it's indeed as you've said: we DO want more more more and it takes more space to do that. But really (and thus the lengthy QNX talk here) these OSes are probably in no danger of becoming bloated and slow by ruling desktop-OS standards. It's just a natural progression, with a little extra "weight" there to serve legacy needs.
