When possible I always capture the document images into PDF files. I store these file in a primary folder called Genealogy. The primary folder has secondary folders for each of the 8 grandparents in my wife and my families. Each of grandparent folders are broken down to documents, census, pictures, burial (Tombstone pictures and cemetery maps),etc. They also have folders for the maternal families with a similar folders for the maternal family information.
When possible, if the site does no provide identification to the file, I will add the information as Comments in the PDF file, or if I can get it as a JPG will add the source information to the border of the image and then print it to a PDF from the image processing program.
When I connect them to my FTM file, I do it in their current locations, not importing them into the FTM system.
(I don't have to figure out where FTM put the image, and it is easier to find it in my folder system.)
I only used web links for temporary connections. I have been working at this for many years, and found the web links either die, or are changed over time andmthen your references become worthless. It does not make any difference if it is Ancestry, FamilySearch, or other, reference links change!
Having all of my data and references on my computer makes accessing the data easier and faster, as I am not dependent on quirks of the provider, web, and IPS connection. Also I am not distracted from my purpose by an new interesting hint or document.