Kyle Jones wrote:
Hrvoje Niksic writes:
> Oscar Figueiredo <Oscar.Figueiredo(a)di.epfl.ch> writes:
> > But I wonder if it's really worth cluttering the code with such a
> > compatibility hack when nobody will actually need it as soon as EUDC
> > an W3 are updated. I agree compatibility *is* a MUST for features
> > that are widely used. I believe the LDAP API currently is not.
> I agree with you. Backward compatibility is a good thing, but it is
> not required in all cases.
This looks like hand-waving to me. We don't know how much the
existing API is being used. That's why we should adopt a
philosophy of never breaking the API once it has been released.
I could understand the "we can probably get away with it" excuse
if maintaining compatibility were difficult. But this is easy.
How are we going to maintain compatibility when it is difficult
to do so, when we won't even insist on it when it is easy?
I would also add that the current configure test for the netscape LDAP
SDK fails because newer netscape SDKs have ldap_open(), but configure
thinks that only the umich ldap libraries have ldap_open(). I've worked
around the problem by running
$ with_umich_ldap=no with_ns_ldap=yes ./configure ....
I'm not sure what the best way to fix this is...