Hi Christof,

Yes, a few have now reported that they obtained equivalent results in 2.6 FPD and FPW.

This strongly suggests that the phenomenon has been there all along, I agree.

Yet I am still perplexed by both all of the prior literature stating rather unequivocally that TOD is beneficial to the point of being mandatory and, most particularly all of those reports that adopting one solved the performance problem previously noted. I think that we all agree that the vast majority of the tables in use (since forever) have only a tiny percentage (if any) of deleted records. I further speculate that most of the people who were reporting problems were *not* using SYS3054 to evaluate their standing vis-a-vis optimization.

In any case, I shall keep my antennae tuned to see if other information pops up somewhere to help to clarify the situation. I remain suspicious , even at this late time.

Cheers,

Jim N

>Hi Jim,
>
>>I rather suspect something else, primarily because of the history behind this issue.
>
>It's not the product, believe me. In contradiction to some people's expectations FoxPro/DOS does not change its behavior because VFP is installed, all it relies on is FoxPro itself and the operating system. Now run the tests yourself on FoxPro 2.6, and you will see that it behaves basically the same. There are some minor differences, for example the way memory variables in a SELECT statement are treated, SET COLLATE has a little more impact, and so on, but DELETED() didn't change.
>
>OTOH, if I look at my code now and my code I wrote 8 years ago, I can see a lot of differences. Not only, of course, do I now know more about FoxPro, but the approach has changed a lot. My early code is full of navigation right on the table. I regularily checked whether a table contained any records, I had a lot of SET RELATION stuff, I used SCATTER/GATHER to read/write to records and of course I re-used deleted records to avoid that my users have to spent an hour or so on packing all tables. After all, I was writing programs that had a total of 250 MB data, quite a bit at that time.
>
>Christof