This is a platform specific recommendation for file-sizing. An example file named DATAFILE is used which can be generally applied using the guidelines set forth below.

1. Assumed DATAFILE file with 100,000 items, each averaging 335 bytes. Requires 33,500,000 bytes (33500000/1024/1024 = 31.95 MB). Also assumed is the need for 100 or more DICT items.
   
   
2. This 32 Megabyte file must be created differently depending on the platform because of the hardware restrictions placed on the PICK/uniVerse environment when reading and writing data to the file.
   
   
3. For example, on Ultimate or other PICK-native machines (except SEQUOIA) you would calculate that the file needs to be created with: CREATE-FILE DATAFILE 29,1 65437,1 or probably slightly larger for a good 'cushion'.
   
   
4. Here are examples using the same file as it should be created on different platforms:
   
   
   1. IBM RT or RS (Motorola, NCR, etc.):
      CREATE-FILE DATAFILE 17,2 32719,2 (or) CREATE-FILE DATAFILE 17,2 16361,4
      
      
   2. HP or DEC:
      CREATE-FILE DATAFILE 7,4 16361,4
      
      
   3. SEQUENT:
      CREATE-FILE DATAFILE 5,8 8191,8 (or) CREATE-FILE DATAFILE 5,8 4091,16
      
      
   4. SEQUOIA:
      CREATE-FILE DATAFILE 5,1 8191,1 (Sequoia's block size is 4 times larger than other PICK-native platforms.)
      
      

These files are different only by separation, as you've noticed (except for Sequoia). It has been discovered that a significantly higher propensity for file breakage occurs when these guidelines are not followed. On IBM, for example, EVERY TIME a file that has a separation of 1 has an item written to the last block, or the last block becomes full, it breaks at that point. When a CLEAR-FILE DATA is performed thereon, the command gives an abort message, but works anyway. Very confusing. If you have any questions, please contact the Operating Systems Support Group in Salt Lake City.