View Single Post
Old 20th April 2002, 00:06   #17  |  Link
Tronic
Registered User
 
Join Date: Mar 2002
Posts: 33
IMO checksums should be in storage media level because that allows easy extraction of broken sectors (ie. format-level wouldn't have checksums for 2332 octet pieces, which would correspond with CD sectors (one type of those).

Also, for additional security, there should be additional checksums in data, ordered so that the data is protected best (in MCF that means that all Clusters should start with video keyframes (a cluster is protected by a single checksum)).

Error-correction is normally not required as data links (TCP/IP) and storages (HDD) are pretty error-free. Even if the harddisk broke or the peer server in TCP/IP died, ECC data wouldn't help us.

Error-correction is even more dependant on the media than checksums are - a CD normally has broken data in 4 octet chunks (audio CD sample size) and have no misaligment (unless we are talking about CD-AUDIO, which always has offset of n*4 octets, I think).

In MCF-CD (Tronic's version) the trick is that the correction is format-independant. My extraction program can extract any file format from MCF-CD-images. Check out MCF-CD header specs for more info.

However, writing programs should be format-specific, so that those know which parts of the file should be protected. It is possible to write unknown fileformats too (even with my tool, with minor tweaking), but there you won't have your headers protected (unless you manually define positions to protect for each file).
Tronic is offline   Reply With Quote