"NSF" files are really comma-separated value files, but because they are horribly formatted, require special processing by Network Workbench. Thus, in order to get Network Workbench to treat them correctly, the extension must be changed to ".nsf".
What is the difference between the "current" and "pre-2014" NSF formats?
A full listing of changes may be found on the JIRA issue page for this update. The most important thing for users to keep in mind is that sets of data are now delimited by commas instead of pipes in the current version. Also, when using aggregate function files, you will want to use specially labelled "pre-2014" properties files when working with those older data sets.