You can:

• Use a perl compiled with 64bit IVs.
• Store the number as an NV, on most CPU architectures, NVs (=C doubles) can represent integers of 48 bits (49 including the sign). For instance, if you are representing file offsets, that means 256 Terabytes.

  SV *get() {
    ...
    return newSVnv((NV)val);
  }
  [download]

  (note that newSVnv() will generate an IV or UV instead of and NV if the data fits inside)
• Use and external module as Math::Int64 or Math::GMP
• Return the int64 as a string, converted to decimal or packed.

Adding a C API to Math::Int64 (I am the author ;-) is in my TODO list. update: DONE!