If you make radical changes to a file, it doesn't hurt to make several backup copies along the way (e.g. "*.bak1, *.bak2 etc.) so as to allow you to go back to an earlier version if you make a real hash of things.
Alternatively: type "sample [filename]" at the command prompt.
At this point, many people print the notes file, attach it as a coversheet to the sample, and use it to write notes on. Others prefer to enter all their comments electronically and do not print out the notes file and attach it till the end of the review process.
NOTE: Of these five categories, only errors in categories (a) and (c) should be counted against the vendor's allowable error rate.
NOTE: when a book is very bad, you may be able to stop proofing early. E.g. if the stripped sample size is 20,000 bytes, the vendor is allowed only one error. If the number of clearly inexcusable errors counted exceeds this (especially if it exceeds it by more than one), you may stop proofing, since the file has already failed. If this happens you should add a note to the notes file of this sort: "stopped proofing after 5 pages because of excessive errors." Some judgment is called for in doing this; we prefer not to reject books without good evidence, and rarely reject on the basis of (say) a couple of missing full stops, or an s/S case error.
NOTE: We usually "give" the book a free error. That is, for small books, we do not reject it for just one error, regardless of size; and even for larger books, we discount one error in deciding whether to reject, accept, or pardon.
NOTE: very small books (books of 5-10 pages), and many slightly larger books (10-25 pages) we will usually pardon regardless of how many errors they have, since we will have effectively proofed the entire book anyway.