[Flac-dev] 32-bit sample size

[Josh Coalson]

--- On Sun, 6/7/09, Justin Ruggles <justin.ruggles at gmx.com> wrote:
> Hi,
> 
> I have a question about 32-bit sample size.  It is obviously not
> supported in encoding or decoding in libFLAC, but is allowed in the
> format description.  The only way to allow it is in the STREAMINFO,
> which means that the files would not be Subset compatible.  The Subset
> constraints list the value of 7 as allowed, even though it is reserved.
> 
> So my real question is:  Is the value of 7 meant to be used for 32-bit
> eventually?  If so, will 32-bit be Subset compatible at that point?

I would probably lean more toward using 3 for 32-bit samples, because
7 could increase the likelyhood of false sync codes occurring if some
of the other reserved bits came into use.  absent a strong demand for
32 bit though, I would prefer to leave as is due to the disruptive
effect of defining new bits.

there is probably no problem with 32-bit being or not being in the
subset.  subset decoders can reject unsupported sample rates; at the
same time I think any application handling 32-bit samples will not
be constrained in a way the subset tries to mitigate.