>>So the theory which says that NOFILTER should force it to exist... actually doesn't come into practice.
>
>No, that theory's been shot down completely by now. The NOFILTER clause does just what it says it'll do, it prevents a filter from being put on a table for the result cursor...

Yes, so now dbf("cursor_alias") returns a tmp file instead of dbf, which still doesn't guarantee the file's existence :).

>Perhaps they should have a new clause (not that it would be very useful) called,
>
>CREATE_DISK_FILE :)

For all the other purposes it works just fine without the physical file, except this shizofrenic division line between datasessions. Maybe another helpful extention could be

Use Whatever again alias NewAlias of Datasession nOtherDataSessionNumber

I've tried all the possible mental gymnastics to open such a cursor-in-the-air from a different datasession, and couldn't find a sure way (short ot copying it into a temp table, but that sort of defeats the idea).