I've had troubles with RTTI that sound like what you're complaining about.

At least the way Borland implements RTTI, you can't have a non-temporary
typeid object (or whatever the class is called), and
can only get one by calling typeinfo() (or whatever the function is
called).
This means you can't (to my knowledge) instantiate objects based on
typeid's,
since the they can't be put into or gotten out of a stream.  Therefore you
can't
implement persistence mechanisms that use typeid's.  I have had to kluge
my
own class id's to get around this.

Regards.

Nate Clark
nnnlllccc@aol.com

> It has already been accepted, and it is *extremely* unlikely that
> it would be removed.

Is it going to be usable to implement portable (among programs
compiled by different compilers) persistence schemes, i.e. is the form
of the typeinfo mandated, or implementation defined.

Brian
[It will be very irritating if we will still have to maintain a code
to map types into names]
--
       Brian Glendenning - National Radio Astronomy Observatory
bglenden@nrao.edu          Charlottesville Va.           (804) 296-0286