Now it seems that I've fallen into the feature trap again. I'm now
hacking to generalize address handling in the socket code. That was
needed for at least three reasons:
* For supporting AF_LOCAL sockets, which will be needed for the
gateway.
* For IPv6 support.
* For logging purposes, that requires that peer addresses are kept
around.
I have also done some work on utmp/wtmp support. I'm starting with
wtmp-logging using logwtmp(), as that seems easiest. (BTW, what is the
Right Way to handle logging of "sessions" that don't have a pty?)
Other items on the list that I'd like to look into, sooner or later,
are
* Encrypted private keys, using the method (but probably not the
format) of PKCS#5.
* Reading of user files (it is somewhat tricky to get right in
lshd, as it will want to read files before it forks and changes
uid.
* SIGWINCH handling.
* Configuration of the userauth policy, as has been discussed here
recently.
* Using ADNS for asyncronous name and address resolution.
Regards,
/Niels
