[Algorithms] CRCs?

[Amir E. <gd...@am...>]

For cumulative CRCs of files is it best to CRC the same bytes with a seed
(from previous CRCs) or is it okay to just CRC all of the file CRCs? If
that's cryptic, then maybe an example:

file1.ext
file2.ext

Would it make any difference between these two in terms of collisions?

file1_hash = CRC(file1.ext);
file2_hash = CRC(file2.ext);
cumulative_hash = CRC(file2.ext, file1_hash);

or

file1_hash = CRC(file1.ext);
file2_hash = CRC(file2.ext);

cumulative_hash = CRC(file1_hash.bytes + file2_hash.bytes);

In other words, does the length of input improve the CRC?

Thanks,
Amir