[bind10-dev] python bindings

[JINMEI Tatuya / 神明達哉]

At Thu, 17 Dec 2009 15:42:13 +0000,
Francis Dupont <fdupont at isc.org> wrote:

> > Yes, we'd like to avoid the in-house approach. Sadly the C++ STL has no
> > hash built in, so may I humbly suggest we use Boost unless there is some
> > compelling reason not to?
> >
> > http://bytes.com/topic/c/answers/630022-stl-hash-function-where-header
> 
> => IMHO jhash is technically better (at least its author is known to
> have done some research about hash functions).

This is mainly about the algorithm.  Avoiding the in-house approach is
more about implementation.

> > This makes sense to me. We shouldn't allow objects to be instantiated
> > that make no sense. :)
> 
> => i.e., names to Name("") (:-)? As I explained for python (or any
> language without overloading or where overloading is considered harmful)
> we need either a "main" constructor or an initial "acceptable" object.
> In current Name C++ class we have no main constructors (two candidates)
> and a buggy initial object: this has to be fixed at the benefit of
> both C++ and python (so now I wait for the new code).

What do you mean by a buggy initial object?  The result of
isc::dns::Name(""); ?  BTW: I also didn't understand what you meant by
a "runtime error" (which you seem to mean is something different from
a DNS exception).  In fact, isc::dns::Name("") will cause an exception
of BadLabelType.

---
JINMEI, Tatuya