[MPlayer-dev-eng] Nut, few ideas

[Ville Saari]

On Tue, Apr 20, 2004 at 06:44:05PM +0200, Michael Niedermayer wrote:

> ok, ill explain it to u again, the checksum must be at the end because when 
> the packet doesnt fit in memory it would require seeking back, it also 
> sometimes simplifies decoding (calculating the crc of the data with the crc 
> checksum appended at the end gives 0 if its undamaged) ok we use the adler32 
> checksum which doesnt have this property ...

Having it at the end also simplifies the specification: you have the data
followed by the checksum for that data. If the checksum is in the middle
of the data, you must also specify how the space for the checksum is
supposed to be handled while the checksum is being computed. Is it
initialized with some constant value or should it be skipped, which
greatly complicates the checksum computation or is the checksum supposed
to be applied to itself using magic?

-- 
 Ville