>>What I further found out:
>>- The maximum size for the stringtype key/info is 256 bytes, I believe.
>>- Each redeclaration (with GlobalAddAtom) of the same atom name increases what might be called a 'stack number'. Each delte (with GlobalDeleteAtom) decreases it. So, the most secure way to get rid of the entry is:

do while GlobalFindAtom( cString ) > 0
>>	GlobalDeleteAtom( nAtom )
>>enddo

>
>It seems to me that this would be an abuse of the original idea of using atoms - which might cause some problems.
>
>What I briefly read some years ago - assuming I read and understood it correctly - is more or less the following.
>
>Atoms are used for efficient storage of strings. Each time you create an atom, with the corresponding functions, Windows will check whether an identical string already exists. If it does, the same space will be used again. If it doesn't, a new atom will be created.
>
>So, it would seem to me that any function must use GlobalAddAtom() to define the string, and a single invocation of GlobalDeleteAtom() to release it. Then, if another program or process happens to still use it, it will remain in memory, as it should.
>
>Reducing the usage count to zero would be an indication that no process, not the one you are currently in, and no other process, uses the specific string any longer.

I agree. Normally, one should use a quite unique string.

Here's a nice piece of code (http://www.news2news.com/vfp/?group=-1&function=167 Anatoliy Mogylevets)

#DEFINE AtomStrLength   512 

DECLARE INTEGER GlobalGetAtomName IN kernel32; 
    INTEGER  nAtom,; 
    STRING @ lpBuffer,; 
    INTEGER  nSize 

CREATE CURSOR cs (atom N(12), strlen N(5), name C(250)) 

* scan string atoms: from 0xC000 to 0xFFFF 
FOR nAtom = 49152 TO 65535 
    lpBuffer = Repli(Chr(0), AtomStrLength) 
    lnResult = GlobalGetAtomName (nAtom, @lpBuffer, AtomStrLength) 

    IF lnResult > 0 
        INSERT INTO cs VALUES (nAtom, lnResult, LEFT(lpBuffer, lnResult)) 
    ENDIF 
ENDFOR 

SELECT cs 
GO TOP 
BROW NORMAL NOWAIT