From: jmose@mapson.attglobal.net   
      
   Anders,   
      
   >> Imagine a 100MB file on network server.  With a 500k index file.   
      
   This was exactly my problem.  One of my clients had a 20MB Customer table,   
   whereas all others have under 1MB.  Their network is dying, and I'm trying   
   to speed it up.  Sorry I didn't post this number earlier, but I didn't have   
   an exact size on the XG1 file.   
      
      
   >How about a  tcursor then? Will a scan of all records   
   >eventually bring the enterie table over the   
   >network, even the columns that is not refered to by the tcursor?   
      
   Sadly, you can't open a tcursor on the XG1 directly - it gives a message   
   'Table is not indexed' (ironically).   
      
   If you open a tcursor on the base table, it will 'pre-fetch' 8 pages based   
   on the block size.  Then, as you jump around the table, it will grab whatever   
   pages are needed.  So, if you scan the entire table, you'll need all 20MB   
   in my above example.   
      
   One other benefit to the XG1:  if you open a tcursor on a table, it will   
   open all table lookups also.  I'm assuming it doesn't want to slow down   
   tc.edit()   
   (or the f9 key), so it always assumes you'll edit.  I'm not sure, but these   
   might also get 8 pages 'pre-fetched'.   
      
   I'm getting somewhat proficient with the free WireShark network monitor.   
    http://www.wireshark.org    I you have speed issues, this is the best tool   
   I've found.  It replaced Ethereal a while back.   
      
   Thanks,   
   Jim Moseley   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)