Well I've to accept I'm terribly mistaken, assuming K&R is saying so - I probably have screwed up big time, despite the fact I've readily available compilers to support what I said. I have not read K&R recently and do not have a copy of it right now, I'll get one from the library. I wonder what the ANSI and ISO are saying. But, beside having to blame the standards, I've the following point to make : If "abcde" is a constant value to be stored in read-only areas, the compiler should stop there and give an error, saying a type mismatch. The case in point is, const char p[] = "abcde"; char *q = p; /* Error. an attempt to assign different typed values */ The same way, if a compiler is smart enough to put "abcde" in rom in the following example, char *p = "abcde"; the compiler should not compile it for the same reason. Because the standard is against what I'm saying, I accept I was and am totally wrong. Thank you for all the information you have given. I do not want to blame the standards. I do not want to say anything b4 I read (this is a big lesson for me), but I wonder what Bjarne Stroustroup has to say.