Code review comment for lp:~vanvugt/mir/output-model

I don't mean fail to parse the EDID on checksum failures, but to expose that the EDID is broken. For some things (particularly modes) a broken EDID *is* worse than no EDID.

Sometimes the checksum fails because the manufacturer failed to compute the checksum correctly. More often it's broken because the EDID, or the DDC pins, or something is actually broken and should be trusted less.

Hm. While I think of it, checking that all the strings we return are printable ascii is probably a sensible thing to do, particularly if we're going to accept known-bad EDIDs