>>>Also, I think hard coding a cursor name is always fine. Cursor names are private to the data session that they are created in and their file names are sys(2015) values implicitly.
>>
>>Here's an exception which made me think twice before giving a fixed name to a cursor... well, until I managed to have every user have his own temp folder. In the bad old times when we didn't know how to do that, having a fixed name for a cursor was fine until that cursor had a cdx and got flushed to disk. Then some other instance created that cursor and fox found the cdx (this is fpd2.6 days) and tried to use it and of course it didn't match - it was made on the records of the first user's cursor. Same thing applied if the app crashed and the cursor's cdx was left on disk.
>>
>>That was 24 years ago, though - so either the defensive practices worked, or having separate temp folders did.
>
>Sorry Dragan,
>I didn't understand the scenario. I create indexes for cursor but do not have the problem you mentioned. Maybe you weren't creating structural index but specifying an index file name.

Yes, I think that was the case. It was long ago, and my.memory.refresh() was finished only after I posted the message :).