There are no QDOS EXECutables in the attached ZIP package, so you are safe to extract the contents on a PC/Mac/*nix before importing on to an SD-Card ready to run on your Q68; you'll only need the single binary file "ndq68_dvr_v305" - unless you also fancy perusing the (heavily commented) source code.
Whilst Peter designed the Q68 with the digital-side of a QL-compatible network port, the Q68 will need a small and simple HW modification to add the required couple of resistors, 3.3v Schottky diode and the 3.5mm Jack socket(s) to allow physical attachment to the QLNET. I'll follow-up this post with some details in the coming weeks, but for the impatient solderers out there, you can find a simple schematic in an earlier post on this forum or else in a PDF available from the German QL Forum here:
http://forum.tlienhard.com/phpBB3/downl ... hp?id=8551
If you do decide to make this mod yourself, do take care not to brick your lovely Q68 - it's very tight in there and remember that it runs at 3.3v (hence the absolute need for the diode)...
The driver (ND-Q68) is based off the latest published QLNET driver source as included with SMSQE/TK2 - and is fully compatible, providing all the same features including the really useful FSERVE network server and client functionality. ND-Q68 has been tested successfully on both SMSQE (v3.34) and Minerva v1.98 (with MK's I/O fix) on the Q68 and can be loaded with a simple LRESPR or equivalent. You should be fine with any recent version of either OS (as available for the Q68.)
Two additional S*Basic FNs have been added - NET% returns the current Network Node number assigned to this host (with the standard NET procedure) and NET_VER$ returns the driver version - "v3.05" as of today's release.
The driver was re-designed to make use of the Q68's 40MHz counter/timer register and thus runs independently of the Q68's (optimised) 68k instruction cycle-timings; all earlier network drivers were devised to be dependent upon the clock and instruction-timings of their respective hosts (QL, SGC, QXL, etc) using simple, software-based loops - and cross-platform compatibility tended to suffer as a result.
In order to ensure consistent timings even on the Q68, the driver automatically copies about 2kB of its code in to the Q68's uncontended 'Fast Memory' during initiaisation - I borrowed an idea from Marcel's TK2 network 'split-code' approach to link the logical and (timing-critical) physical segments, albeit in a slightly different and extended way. Take care then if you already use the Fast Memory area of the Q68 as it may clash with this driver.
One outcome of the use of the constant 40MHz timer-tick and the Q68's uncontended SRAM is that the network timings more precisely match TT's original specifications and, as a result, ND-Q68 proves to be more tolerant of the timing vagaries of other connected QLNET nodes. In fact, ND-Q68 can even communicate reliably in both directions with a Spectrum/Interface-1 - as far as I'm aware, something not seen before between QL and Spectrum
ND-Q68 was successfully tested alongside QLs (with TK2 in ROM or uncontended RAM), QL+SGC (Tetroid) and a QXL. The Q68 was able to communicate with nodes that couldn't reliably talk to one-another. Aside from the network 'Broadcast' function (which was re-designed for TK2), you should find that even without TK2 loaded, a Minerva or QDOS equipped QL will communicate with the Q68 at a basic level too.
Furthermore, given the Q68's reliable, high-speed SERial port, Peter has also already had some success running ND-Q68 alongside the brilliant SERNET driver, allowing the Q68 to act as a 'bridge' between those nodes with real QLNET ports and other machines such as the Q40/Q60 and emulated QLs running at up to 115,200 Baud.
I've got further plans for ND-Q68 (v3.06 is already brewing...) as well as the respective network drivers on our humble QLs and emulators, but would welcome feedback on use of ND-Q68 in the 'real-world' in the meantime.
May I offer my personal thanks to Peter for bringing us the Q68, Wolfgang for his patient tuition and guidance on assembling from the SMSQE source code and Jan B for Minerva4Q68. Of course, one mustn't forget the original genius of TT, the QView team and all those involved in developing what was and is a truly effective and robust network capability (with the right hardware, at least) - way back in the mid-eighties and all in less than 6kB!
I'm happy to field and attempt to answer questions about the driver here, but do remember that in true open-source fashion, you have the source code available to study and modify/improve upon as you see fit - it was a labour of love, but it's yours now
Happy QL-Networking!
