[Vorbis-dev] Again: Next libvorbis release?

[xiphmont at xiph.org]

Max wrote:

> While doing the conversion, I noticed a discrepancy between the
> formulas in section 6 (on floor0): the PNG version of these formulars
> differ from the LaTeX code that was embedded in the XML code. Now I
> wonder which of these are correct. For now I put both variants in
> there, in section 6.2.3, page 41 of the PDF (the different variable
> names are pretty minor I guess; but the changed indices on the
> coefficients probably are not).

Ralph wrote:
> So I looked at this for a while this afternoon. I think I have the
> coefficient indices correct, but I couldn't reproduce the factor of
> '4' inside the products, nor the variation in the prefactor
> denominators.

The LaTeX code is wrong, the PNGs are correct.

The prescaling factors do match the code, they're just 'hidden' in the
code by being moved into other steps.  The .5 factor in front falls
out of the final squaring of p and q which are initialized to
.707[...]. The difference in prescale in the odd version falls out of
the odd LSP computation algebra. Scaling the odd terms only appears to
be different, the equation makes it difficult to see what is actually
happening logically). The 4 factor on the cosine subtraction is
handled later by adding 'm' (the order) to the exponent after the
product is normalized.

Max wrote:
> Another thing some people might find annoying are the boxes that
> surround all hyperlinks in the PDF. I can turn this off, or
> alternatively, make the link *text* colored. The colors themselves are
> also customizable. Just tell me what you prefer.

Yes please!  I'm used to active links in PDFs being blue text,
possibly underlined.

Monty