[lucianbr]

A hash table makes sense in memory. If it's loaded just right for fast access, it has holes - empty entries. That makes little sense if you are building a file that will be transferred to many places over the internet. Bandwith waste would be significant.

So it might seem that simply enumerating the data (sorted or not) would be a better option for a file. (After all, the receiver will read everything anyway.) I guess frequent updates make this inefficient, so a tree helps.