We've encountered situations where PDF files are not always handled as binary files in the source view of a repository. Instead of seeing "Download to view this file", we see the source of the file, including unreadable binary streams embedded in the file. It appears as though BSERV is only checking a subset of the lines in the file for binary data, so larger files with binary streams towards the end may be incorrectly identified as text/source.

PDF should always be considered a binary file.