@ bloodline:
I am familiar with little-endian/big-endian issues since Z80 and 6502 era. It is a difficult problem (if not impossible) to resolve 100% at the binary level when transitioning to a big-endian platform.
Let's think about how critical or non-critical the interaction of 68k with x86 world really is. If you have compiler (or compilers) which can target one or the other environment, there is no issue. If single binary should run in both environments, we have to talk about universal binaries (another term from Apple's vocabulary).
If you want to write software that crosses the 68k-x86 boundary (for whatever reason), you'd have to integrate these applications as if they lived in two different memory spaces, two different CPUS, two different of everything. We are talking pipes, web services and god knows what else here. New APIs. New Library.
How many applications would do this? Probably none, and certainly no existing application running on Amiga OS would do it (they are binaries and don't know any better
.
We need 68k mainly for compatibility. The x86 environment is supposed to attract developers familiar with Amiga OS and APIs to develop *new* applications, while being able to run the old ones. Initially, development tools *for AROS* would be needed to jumpstart this. 68K environment also has limitations when it comes to 32 vs. 64-bit and this is another reason why it should be contained in its own "Classic" environment. Future platform must not be constrained with 32 bit.
What you pointed out about Apple and selected few killer apps is absolutely true. Mac OS X started out with very few apps but they were all very good. Apple also had Microsoft's support (with Office) and that means a lot in marketing terms.
Even if all these pieces were working *today*, I very much doubt that any corporate entity would embrace Amiga. Too many companies underestimated/misunderstood/burned/got-burned-with the platform that anyone would think twice before investing a ton of money into Amiga. This is the strongest argument for open source effort. With a good single video driver (video drivers are actually very good even today. You should take a look at "Doom" shipped with AROS) and single good audio driver, AROS would become a viable development platform.
Again, Apple is a good model here: they bake video and audio onto the motherboard and that's it. No hassle with umpteen drivers. When you think about it, Windows is the same; it works flawlessly on Lenovo laptops (for example) and falls on its face on cheap clones with questionable components.
68k platform is history. It is a 30-year old technology and it is a dead-end. PowerPC is also a dead-end in terms of future development. AMD and Intel are the only two companies committed to cranking out new hardware (boards and CPUS and chipsets) that you can rely on. The "only" thing standing between Amiga OS and a-hardware platform are suitable drivers and that is the most difficult part. Good news is that most popular pieces of hardware tend to be well documented (in terms of what they support and what the issues are).
Once everything is working, the question becomes "how is this any different than Mac or Windows?" Amiga would have to have siginificantly better bandwidth for graphics/video intensive development *and* multicore support to establish itself. As of today, Amiga OS does not support multiple cores. Starting next year, Apple will push OS X much stronger into multicore market with the new improvements.
Also, it would have to have better development tools and support from major software manufacturers (from games onward). It is a fairly tall order.
Last 20 years brought about a significant change to the marketplace: even high-end hardware is a commodity. However, as Apple showed, *making it all work together* isn't a commodity.
I keep mentioning Apple because they almost killed off Mac platform in the early 90s. Then Stevie came back and made it really sexy with flashy plastic and same crappy OS. Everyone bought into it. While everyone was busy admiring colorful plastic, Stevie was working on OS X.
Heck, first thing we need is some flashy case for the Amiga. Titanium-alloy thing that looks like it is moving at fast speeds. With only one red LED in the middle. If you put an Amiga back into the beige case, might as well trash it immediately.
Quote from bloodline:
68k is not a problem as long as the 68k programs are not allowed to interact freely with the x86 programs. That is why we want to take the approach of putting the 68k apps into their own little VM world of UAE and carefully control how much interaction they have with x86 AROS.
If we simply integrated the 68k emulator into the OS like MOS and AOS4, then the 68k apps would be allowed to "play around" with the x86 system structures... the problem is that on x86 data is stored : "DCBA" (little endien) and on the 68K the same data is stored: "ABCD" (big endien)... so 68k apps in a x86 environment will read the data the wrong way around, and thus fail.
Things get even more problematic when you use the 64bit x86 verion of AROS (which is probably more advanced now than the 32bit x86 version).. where the 68k apps have no concept of 64bits at all (no 68k Amiga software is 64bit compatible) and wouldn't work even if the x86 was big endien!
All my x86 machines are 64bit now...
