Hello,
I have three questions, though all of which already mentioned by other people in
the mailing list,
about the accuracy of the Vorbis I Specification published on Feb 27.
1. On page 34 section 4.3.4., explanation of the situation where there are only
one submap is omitted,
and it is difficult to understand that the decoder should assume in such cases
that all the audio channels belong to submap #0 in the original order without
examining the array [vorbis_mapping_mux] (which is empty).
2. On page 58 section 8.5., it is unclear as to which of the following
implementation is correct:
? A. For each partition, decode a vector the size of [partition_size] * [ch]
according to Format 1 and deinterleave. Repeat the process over all partitions.
(wrong)
? B. Decode a single vector the size of [blocksize / 2] * [ch] according to
Format 1 and deinterleave once. (correct)
3. On page 60 section 8.6.2., taking the greater of [residue_begin] and
[actual_size] will almost always produce [actual_size], which does not make
sense.
? Taking the lesser of [residue_begin] and [actual_size] as
[limit_residue_begin], and the lesser of [residue_end] and [actual_size] as
[limit_residue_end] seems correct.
I came across the questions above in the course of writing a new decoder which
is almost working properly,
(https://github.com/sohc1300/Soyogi)
and I am quite certain that it would be very nice if these points are clarified
by official documentation.
Please excuse me if there are any misuse of English, as I am not used to.
Regards,
Hiroka
--
Hiroka Ihara
Department of Information and Communication Engineering, the University of Tokyo
PC: ihara_h at live.jp
