Compression and file integrity

[rangopango]

Is it true there is a chance to damage the file if you transfer it to a dataset with enabled compression?

 

[garm]

Said who?

 

[Chris Moore]

Is it true there is a chance to damage the file if you transfer it to a dataset with enabled compression?

How would that happen?

 

[rangopango]

When you compress a file, such as a song, there is a chance for loss of information. Doesn't the same thing apply for file system compression

 

[Chris Moore]

When you compress a file, such as a song, there is a chance for loss of information. Doesn't the same thing apply for file system compression

If the file is able to be compressed, it will be, if not, then it isn't. ZFS is all about giving you back exactly the data you stored. The music and video editing software use an algorithm that throws data away to make a smaller file. That is not how actual compression works.

 

[danb35]

When you compress a file, such as a song, there is a chance for loss of information.

Um, no. When you compress a song or a video, you're typically using a lossy codec. There isn't a "chance" of losing information, there's a guarantee that you will lose information--it's designed that way. But when you .zip a file, for example (or, really, do any other kind of compression), you're using a lossless technique. There isn't nearly as much space saving available, but the output is identical to the input.

 

[toadman]

Yep. Perfect example is using h264 (or other) to compress a 30 GB Bluray. You can get it down to <2 GB and it's still watchable. That compression algorithm is lossy and designed to take away detail that "won't be noticed if it's gone or smoothed out" as a trade-off for file size (bit rate).

Contrast that with LZ4: https://github.com/lz4/lz4

LZ4 is lossless and specifically trades off compression level (vs say gzip) for speed.