You're confusing extraction with mixing. Extraction is already being used in the matrix decoding modes, so allowing it to be variable is easy.

Sigh. Unless said extraction is non-linear, full extraction followed by variable mixing can have the same effect, and allows leverage of an existing mixer.

There is no mode that uses variable bleeding of discrete centre content. Since it's not there to begin with, how can it be restricted to certain modes?

All the pieces are there: extraction, and mixing (how else would one mix to the appropriate number of channels, and implement the variable balance between side and back surround channels if there was not a continuously variable mixer?).

If a variable extraction to centre is available, a fixed extraction to centre is available, and since a mixer exists, it can be used whether the centre is discrete or extracted. It's a simple issue of software refactoring.

Outlaw can't change how those modes work (even if they knew how to) because those are licensed items and as such have to adhere to the specification of the license.

No, but they can refactor bits out of the implementations! I doubt the licenses prohibit this -- they likely include a reference implementation and a suite of tests that the implementation must pass. What is licensed is not likely to be code specific to the processor in the 990, but rather a right to implemented a patented process, along with access to intellectual property to facilitate such an implementation.

Also, it doesn't add any complication to the documentation of the modes. Different modes have different features. You just have to deal with it.

Have you never heard of the benefits of software refactoring? The common aspects of different modes can be refactored. In fact, I'd bet that, in the 990, they are, but that this is hidden from, the user.

If you can refactor a mixer, you can expose it fully, and common mixes to specific number of speakers just becomes a special case of a generic parametrization. The questions are only whether that generic access and the refactoring is exposed via the user interface or if the unit is crippled.

There isn't a mixer present in the 990.

And I say, based on what the 990 does, there is. It just isn't fully exposed. I can't believe that the 990 would used fixed hardware mixing after all that software decoding (which is more computationally intense).

The triple-mono result induces comb-filtering and smears localization.

Well, this is a sensible argument. But you'd get that with partial PLII centre extraction, no?

Actually, with the BG Radia 220i, you do get smearing of localization: the damn thing is more than 32" wide and transmits along its length with a cylindrical dispersion pattern. Centre appears to come from the display, not the centre of the display. You don't get the comb filtering problems with the one speaker alone, but it is possible with a bad (i.e. relatively equal) bleed between separate L, C, and R.

Now it's "hackers"? (I feel like I'm in a sitcom but just don't know it yet.) What can I say, except that Outlaw and Dolby and DTS are not manufacturing products designed with software hacking in mind.

Well, they should. They're already catering to early adopters who think nothing of dropping $1100 to buy something over the Internet. They'd get feature and bug fix prototyping for free.
_________________________
no good deed goes unpunished