🚀 go-pugleaf

Is strtok()'s behaviour defined for tokens already taken from a string?

For example, suppose I have:
    char zoe[] = "aaa bbb";
    char *first, *second;
If I do:
    first = strtok (zoe, " "); 
    second = strtok (NULL, " ");
    printf ("%s\n%s\n", first, second);
I expect the output:
    aaa
    bbb

First idle query:

Can I *portably* change the contents of a token returned by strtok
without affecting the behaviour of subsequent calls?
If I do:
    first = strtok (zoe, " "); 
    printf ("%s\n", first);
    strcpy (first, "!!!");		/* corrupt the first bit */
    second = strtok (NULL, " ");
    printf ("%s\n", second);

Do I have a guarantee that I'll get the same output?
What if I change the 'corrupt ...' line to
    memset (first, 0, strlen (first) + 1);

Second idle query:

Does strtok() promise that the tokens delivered will be untouched
by subsequent calls to strtok()? Could a conforming implementation
keep the original pointer to the string to be tokenised and use the
string itself to retain state (say by filling from the start of the string
through to the last token delivered with some absurd character)?

I've never met an implementation which did care about previously
delivered tokens or which went back and wantonly trashed them, but I've
never seen any documentation which says that a perverse implementor
couldn't choose to do this.

--
Jim Segrave (Segrave Software Services)     jes@grendel.demon.co.uk