[VCDImager] 2 feature-requests/suggestions

[Siegfried Hildebrand]

Hello list!

I have some suggestions about sdditional features.
[warning! this is going to be very theoretical... ;-) ]

1. Seamless playing chapters:
Let's see if seamless playing chapters are possible. (I am still hoping)
We all want to have one MPEG-Track with many chapters on it, to make
winding all over the film AND jumping forward or backwards to specific
chapters possible (like the DVD ;-)
In theory this would be possible, if we set a start- and end-TIME in a
menu-entry in the PSD.SVD. But all players skip the last ~2sec when
they jump to the next menu-entry (= next chapter). Well, indeed this
happens when making menues with I-Author - I don't know if 
vcdimager already fixes this...
If the above doesn't work with vcdimager correctly, too, I have an other 
suggestion wich *could* possibly work:
- we have again our 1-MPEG-stream-track.
- set TRACK-start- and end-points according to the chapter-starts and
   -endings
- now we have _virtually_ multible tracks with NO gaps between them
   on the SVCD. (btw I-Author & Nero always inserts some empty 
   sectors when making multible tracks)
- set the menu-items to play the [virtual] tracks from start to end,
   and at the end: jump to the next track (or PREV: to the previous track) 

Well, we would have a very similar situation as with the common way 
(making menu-items with start- and end-TIMES), but I have some 
hope that my suggestion could work, because no TIMES are set in the 
PSD.SVD wich could be misinterpreted (or falsely rounded in some way). 
Perhaps the firmwares could deal with such a SVCD in the way we like 
it to...
Has anyone got some ideas on this?
Has anyone already made or seen a SVCD with working seamless-
playing chapters?

2. Some extra CD-space-saving:
I-Author SVCD-images are known to produce very few playback-
problems on standalones. Surprisingly the I-Author muxer doesn't 
add the SVCD scan offsets nor does it align the sequence-headers.
Apparently the players (firmwares) don't really need them, if they get 
a correct SEARCH.DAT...
Since (I think) everybody wants to save every bit on the CD, it would
be a good thing to take these 18bytes per I-frame out of the stream.

Much more bytes are wasted by sequence-header-aligning. Which 
means that padding packets are added.
My (and many other people's) streams play fine without that padding.
The only thing I encountered is, that when jumping into a MPEG-Track 
(via PSD.SVD) very seldom the sequence-header doesn't seem to 
be found and the FIELDs are played back in swapped order 
(of a progressive PAL MPEG2-stream!). I know this sounds really 
weired ;-)  ... If I press PAUSE and PLAY the problem is gone.
So, sequence-header-aligning seems to be necessary _only_ at 
menu-jumpIn-points. 
One possible implementation could be to mux the stream without
aligning & scan offsets and when writing the image (when 
vcdimager *knows* what position is a menue-jump-in-point) do
the padding of the particular 'jump-in'-pack.
I think this extra saved space is worth the effort isn't it?

I know that my suggestions possibly violate the IEC-62107, but
since SVCDs aren't made for eternity, and my player (for now)
plays my slighty-out-of-the-spec-SVCDs fine, why not save some
extra bits ;-)
(Btw. could someone send me a copy of the IEC-62107?)

Let me know, what you think about my suggestions.

Greetings,
Siegfried Hildebrand
address@hidden