(2012-05-01 19:04)bornagainpengui Wrote:(2012-04-30 09:27)welshblob Wrote: Given that these boxes are being manufactured already and new ones based on this chip seem to be released weekly then is there a need for a kickstarter project?
Because it's not about the hardware, it's about some sort of standardization and reference design. Think Boxee Box without the lock-in, then add Android games.
Right now everyone just kind of builds something or starts with an old pc sitting in the corner and installs XBMC. It works. It works better than any of us have any right to expect it to work, a testament to the ability of the developers here--but it never works 100% of the time due to strange hardware quirks and differences in configuration.
My thinking (as an enduser, not a programmer) is that if one set of hardware was chosen, then it could be targeted towards much more easily because the developers would have a common platform on which to work. Instead of trying to build something and hoping it will be "good enough" or buying a customer device like the AppleTV and hoping it can be hacked to doing what we want this would allow people to buy a finished box and plug it in. This would be a better deal for both group, those who want to tinker and those who want it to just work.
Like I said, I'm not a programmer so I'm probably missing all sorts of reasons why this is impractical and not likely to work. To me though this kind of thing just makes sense and I really don't understand why nothing like it hasn't happened already. I understand the reasoning behind not wanting to make a XBMC Wii, or an XBMC 360, or PS3 edition because there is no way to prevent the host company from wiping out your efforts with a firmware update. I've never understood why the XBMC team never sat down, talked amongst themselves and said "Hardware XYZ is what we will target and support with OS ABC and the amount of RAM 123, etc etc. XBMC may work on other hardware but this is what we recommend."
While the idea of a standardized hardware platform for XBMC is appealing, it's just is not practical while working with in the XBMC construct. The primary reason for that is because it is slightly in conflict to one of the core goals of XBMC. XBMC was/is designed to provide the best cross platform media center that exists, and to that end, there is no doubt in my mind that our talented developers have done it. Additionally, much of the base functionality (UI, database functions, etc, ect.) have already been or are in the process of being coded such that they preform equally well on all platforms (ex: compatibility with both GL and GL-ES). Really, if you get right down to it, XBMC functions very similarly on all platforms.
The most prominent difference that people notice (and often make a fuss over) is of course with how XBMC decodes video files. On systems that don't support anything fancy decoding is all done in software on the CPU. However, if your system has supported hardware, the decoding process can be offloaded onto that hardware (for example your GPU or hardware decoder card). The XBMC team is/has-been hard at work improving support for various hardware decoders as API's become available (VDPAU, VAAPI, DxVA, Crystal HD, etc.). A good example of this is with the Raspberry Pi and its support for hardware H.264 and MPEG4 decoding(The API Broadcom has made accessible to Pi users grant access to that specific hardware decoding functionality). The beauty of this type of modular design is that it becomes "relativity simple" to add support for additional hardware decoding features as the access to the hardware features become available through the manufactures API.
With that said, It would seem at-least in my mind that the A10 looks to be an extremely promising as others have noted, as it would appear - there is both an android and linux API available giving full access to the hardware decoding features supported by the SoC.