from http://bugs.musicbrainz.org/ticket/5791

Maybe libdiscid coders here could give a hand and write a more low-level-brutal-aggressive-paranoïac ISRC scanning algorithm because the current simple read method often results in mistakes with used discs (I use Windows version).

The result is slide down duplicate ISRC bug™.

Here is a sample of the bug, imagine the following imaginary ISRC tracklist:

1. JPKI00000001
2. JPKI00000002
3. JPKI00000003
4. JPKI00000004
5. JPKI00000005
6. JPKI00000006
7. JPKI00000007
8. JPKI00000008
9. JPKI00000009
10. JPKI00000010

Now imagine this disc has slight glitches or scratches somewhere around the middle and end of disc... you might obtain the following with isrcsubmit:

1. JPKI00000001
2. JPKI00000002
3. JPKI00000003
4. JPKI00000004
5. JPKI00000005
6. JPKI00000005 instead of JPKI00000006
7. JPKI00000007
8. JPKI00000008
9. JPKI00000009
10. JPKI00000009 instead of JPKI00000010

You see, it takes the previous one.

Well sometimes, you are using the cd drive at the same time (which is bad) and you get 000000000000 for those tracks isrcsubmit was scanning at the same moment.

See this forum post for details on what other programs have the same bug (EAC, Feurio!) and what programs don't (Tau Aanalyzer, mediatools) including an implementation that works (mediatools by simonf).

See also some examples of wrong dubmissions because of this bug.