// *** before: ***
// string meant AnsiString: one byte per character, and the active code
// page decided which letters existed at all
var
  S: string;        // AnsiString
  C: Char;          // AnsiChar - exactly one byte
begin
  S := 'Hello';
  Writeln(Length(S));        // 5 - characters and bytes were the same thing
  Writeln(SizeOf(Char));     // 1
  C := S[1];
  // WideString was the only Unicode string type, and it was a COM BSTR:
  // no reference counting, every assignment copied the whole text
end;

// *** in version 2009: ***
// string means UnicodeString: UTF-16, reference counted, copy on write
var
  S: string;          // UnicodeString
  C: Char;            // WideChar - two bytes
  A: AnsiString;      // still here, and it now carries a code page
  R: RawByteString;   // bytes kept as they are, with no conversion
begin
  S := 'Привет';
  Writeln(Length(S));                 // 6 characters
  Writeln(Length(S) * SizeOf(Char));  // 12 bytes - Char is no longer a byte
  C := S[1];

  A := AnsiString(S);                 // conversions became explicit
  R := RawByteString(A);

  // the rule that broke old code: Length counts characters, never bytes,
  // so every Move, ReadBuffer and SizeOf calculation had to be revisited
end;