This is a slight rephrasing of a post to comp.lang.c++, in a thread
"ostream derivation - advice please." -- I didn't get an answer
there, but it's a standards question, so I thought I'd have a go here.

In article <jss-0912940002330001@166.90.100.8>,
Jerry Schwarz <jss@declarative.com> wrote:
[with regard to using xalloc and iword for misuse-checking in paired
 sets of manipulators]
>Two refinements:
[The second refinement was: be paranoid about iword initialization,
 regardless of what the current working paper says]
>
>1) Use a function to access the index value.  This avoids order of
>   initialization problems if you use mystream's in constructors
>   of global objects.
>
>  class mystream {
>     public:
>      static int myFpind() ;
>      ...
>   }
>
>   int mystream::myFpind() {
>      static int ix;
>      if ( !ix ) ix = ios::xalloc() ;
>      return ix;
>   }

It's clear that this technique requires ios::xalloc not to return 0 as
a valid index. If it did, mystream::myFpind could return 0 on its
first call and a different value on subsequent calls.

ios::xalloc in the iostream library I have (came with our '92 vintage
IBM xlC) returns 0 as the first index.
Is this incorrect according to the _current_ state of the working
paper? Or did Jerry slip up?

Cheers,

Ian.