File schema

{"camliVersion": 1,
 "camliType": "file",

  //
  // INCLUDE ALL REQUIRED & ANY OPTIONAL FIELDS FROM file-common.txt
  //

  // Required:
  // (redundant with sum of contentParts sizes, but useful.  if different,
  //  this value is canonical and clients should either truncate the file
  //  at this point or pad the rest with zeroes, as if there was a missing
  //  sparse file segment missing from contentParts)
  "size": 6001034,  // i.e. 1024 + 5000000 + 1000000 + 10 from below

  // Optional, if linkcount > 1, for representing hardlinks properly.
  "inodeRef": "digalg-blobref",   // to "inode" blobref, when linkcount > 1

  // Required.  Array of contiguous regions of bytes.  Zero or more elements.
  // Typically will be just one.
  //
  // Each element must have:
  //    "size": the number of bytes that this element contributes to this file
  //
  // Optional:
  //    "blobRef": where to get the bytes from.  if missing, the bytes are all
  //               zero (e.g. a sparse file hole)
  //    "offset": the number of bytes into blobRef to skip.
  "contentParts": [
    {"blobRef": "digalg-blobref", "size": 1024},
    {"blobRef": "digalg-blobref", "size": 5000000, "offset": 492 },
    {"size": 1000000},
    {"blobRef": "digalg-blobref", "size": 10},
   ]
}

// TODO: Mac/NTFS-style resource forks?  perhaps just a "streams"
// array of recursive file objects?