>>George,
>>
>>Novell or not, if I have a file that is 10MB in size, then I pack it and it becomes 8MB in size. When I add more records it will become more fragmented. If I leave it 10MB and reuse exisitng records in it it does not become more fragmented.
>
>Hi Jim,
>
>I understand that. There are circumstances, however, where it becomes either a minor point. Since I deal with both...< g >
>
>Take the instance of relatively small tables. On the network I work with the cluster size is 512K. Any table smaller than that can't fragment. So access can't be slowed. I deal with quite a number of tables that are less than 100Kb.
>
>The other case is where very few deletes occur, and it's next to impossible to determine what, if any, optimal table size. In this case, tables with production records come to mind. These tables are added to daily with very few, if any deletes occuring. There's little, if any, reason to try to "re-cycle" since the deletes are few and far between. Further, since all the reports are produced by queries, having to exclude deleted records would serve to slow the reporting process. However, because deletes can happen, I have to account for them, and thus use PACK.

Wow... 512K is a HUGE cluster size. Are you sure that's right? Most networks I've worked with (NT or Novell) use cluster sizes ranging from 512 bytes to 16K bytes.