[MPlayer-users] Re: divx 6

[Rich Felker]

On Fri, May 05, 2006 at 09:01:56PM -0400, Jason Tackaberry wrote:
> On Fri, 2006-05-05 at 21:04 -0400, Rich Felker wrote:
> > I was talking exclusively about OGM here. Matroska has significant
> > problems, but on the outside "works ok" for many uses.
> 
> Academically curious: what problems?

- High overhead unless you use "lacing" (which omits a lot of
  information).
- Massively excessive complexity. Implementations are HUGE and that's
  just for the bitstream parsing/writing, not even the semantics.
- Lack of time base (all timestamps are stored in common approximate
  power-of-10 units which is lossy and causes high overhead).
- Block-based windows-type design rather than bytestream-based
  unix-type (and mpeg-type) design.
- Cannot write a fully proper file without seeking (nonstreamable).
- Precise seeking without index is difficult (not supported at all in
  some implementations it seems).
- Does not address the issue of storing multiple streams (e.g. both
  audio and video) with infrequent unaligned I-frames, needed for
  advanced audio compression. In theory it's probably possible but
  it's not addressed and will most likely cause problems seeking.

And probably more I'm forgetting...

Originally I flamed Matroska a LOT when it was introduced, because I
considered it very poorly thought out, designed with a windows
mindset, designed by people without actual experience with the
requirements for a container, focusing on gimmicks and features rather
than solving the core problems, etc. Since then I've come to like it
somewhat due to the complete and utter badness of other containers
that were threatening to become popular (OGM and MP4), and due to the
fact that mkv files I get 'just work' most of the time. However
technically speaking it's still quite lacking, and...

> And does NUT address these areas (if applicable)?

...yes, NUT is much better. :)

Rich