As soon as we deal with this, we also MUST deal with binary files internally, and we MUST mark them as such so they are consistently kept binary, no matter what.

Getting back to line endings, there are a few things that aren't quite right:

1. the internal line ending should be constant, the user should not be able to change it via get_linesep_conv as it is today.  All that should be possible for the user to configure is what should be regarded as the line ending in text files in the workspace.
2. monotone currently adds a line ending at the end of a text if it's missing.  This is because split_into_lines() currently has no semantics at all to show that, and join_lines() assumes that every line should end with a line ending, even the last one.  This is easy to fix in itself, but it affects everything that uses split_into_lines(), and there are quite a number of routines that do exactly that.
3. monotone currently splits lines at ANY known line ending, so it will split at all CR, LF and CRLF.  I'm not at all sure this is bad, but it's being discussed quite heavily on the mailing list.

/Richard