Hi Sean,
Actually, I think you have some very good points but I'll add a few more (well, of things I hope represent good observations/points)... You are absolutely correct that a proprietary control is typically one which immediately suffers from the age old problem of being almost impossible to update. That is one area where a PC based system excels. To be clear, although I don't care much for Windows being the control OS, I don't have an issue with the hardware; clearly is is very reliable and robust or it wouldn't be used in everything from ATMs to running hydro plants (ATMS historically ran OS/2 for a long time, but the hardware is not much different from the normal PC... all from what I have been told).
However, MS Windows is not (IMHO) the best platform for something like a CNC. Personally, I'm much more partial to Linux there. That puts you squarely in the land of EMC/LinuxCNC as the most viable way to go. Ironically, that EMC is actually the core of Mach3, something that many do not realize. Art took the publicly available software and wrapped it with a UI layer (among other changes) which became the nexus of Mach. We are now in the post-Art world and the new owners of the application claim that Mach4 is new code from the ground up... We'll see. Anyhow, I think I've digressed from the topic a bit.... Mach3, and probably Mach4, do not work in a real time kernel because Windows is not real time. Mach gets around the problem by installing a very low level, predictable, interrupt on the PC; think of this as a heartbeat. Linux is available as a real time kernel for free, without any games played to get it to behave properly.
The biggest consideration with Mach, and the main reason why I was steering away from it, however, was that it is limited to working with steppers. Now, there is a huge debate of stepper versus servo, to be sure! I don't want to step (pun) into that foray, but I do strongly feel that I'd rather have real feedback than perceived feedback on something as large as a Deckel; this ain't no Sherline (not to disparage that machine as I think it has a great, and well deserved, following). People get around the stepper issue by using drives such as Geckos. They take step/direction and control a servo; but the fundamental truth is that the Mach is still stepper based...
The step/direction thing plays a huge role in a conversion of a machine if you want to maximize component reuse. If you run Mach with a Deckel, you'll need to install new amplifiers at a minimum. That is a bit of a downer, IMHO, and really causes me to want to put another nail into the Mach coffin.
The next nail is reputation. Mach has a wonderful following but for the last several years there have been numerous complaints about stability and supportability. I'm not entirely sure what happened there but I'm guessing that Art's moving out of Mach was either symptomatic or a result of these complaints. But this does beg the question of "who do you call when nobody answers the phone." Then again, as you have pointed out Sean, if the hardware is generic then you can just switch.... assuming you don't have a problem with stepper vs servo subsystems.
All in all, my original idea was actually to use Mach. However, this was to be a learning experience only. I had always intended on preserving the same feel of the Deckel... Basically I wanted to build a Dialog 4Plus. After completing that, my intention was always to keep the UI module as a generic package so that I could shift it over to EMC. Items such as the logic for the drive shifting were likely going to be handled in a separate micro-controller sub-system, etc.
I haven't given up on the conversion idea, actually... I know Don feels Deckels are dead, and to be honest, I'm not sure I believe there is a 10M market out there either, but I'm a geek. I love working in the machine/human/software world and it is just a plain old interesting challenge to me! That said, the test subject, Deckel-stein, travelled across the river Styx, as you so personally know. It is ironic that I'd been getting pretty excited about taking this all on. In August I ripped the guts out of my Hurco and was moving that to a 5 axis LinuxCNC control. This was my test subject and it is moving along pretty well.... all a different story. So, I'm not done yet with the Deckel CNC conversion idea. I may get the chance to do it still... Sadly, I'll probably never get a stab at the thing that makes the Deckel so nice, though (the all angle table), even if I ever get another machine.
Alan