[openbeosnetteam] Re: Some quick questions

translation of Waldemar's mail follows:

On 5/30/06, Axel DÃrfler <axeld@xxxxxxxxxxxxxxxx> wrote:
> > Ah, ok, I'll try to research about the legacy headers then, maybe
> > sysctl is
> > defined somewhere?
>
> I only know that the private networking headers have some sysctl-stuff
> in them. I would rather see it go away, though, and have a network
> specific configuration method via ioctl()s only.
> Also note that BeOS BONE does not have sysctl. If you ask me, I would
> like to put the whole network into a BeOS-like shape with useful
> defaults and none of the usual config files left (/etc/protocols, /etc/
> networks, /etc/services, ...).
> I would like to have an application called net_server (or
> network_server) that brings up the networking stack, configures it, and
> then plays the role of inetd - with configuration via BMessages that
> are done immediately without the need for a restart (of that server).

And I would also like to have a method to create dynamic DNS entries
for DHCP and PPP connections that are guaranteed to go away after a
crash. Actually, I'd like to have an easy method for creating DNS
entries, in general (i.e.: without editing text files). :)
Do we have enough man-power to get this finished? We certainly need
better communication methods (BMesssage) and a cleaner desktop
networking approach than what FreeBSD can give us out of the box.
Active developers:
* Oliver?
* Axel?
* the secret paid developer I don't know the name of ;)
* who else?

> pselect() is exported from libbind.so under BONE, but maybe it's not in
> R5, and that's why Waldemar removed it - since he missed the
> opportunity to write down his reasoning, all we can do now is publish
> it again and see where it breaks :-)

Hey! :)
No, I removed it because of build errors. I think there were two
conflicting declarations when compiling for BONE.
I admit I should have mentioned the exact build problem.

> I think the DEVNOTES approach as is, is not very useful - the status
> quo should be retrieved from the source files anyway (as such docs are
> always out of date). If you don't write your reasonings into commit
> messages, I think that what's belong there: a short overview over the
> different targets (userland vs. kernel), and reasons on why things have
> been done the way they are.

What do you mean with getting it from the source files? How do you
want to do that (comments, diffs, ...)?
It should be easy to reconstruct our adaptions if we want to keep our
port in sync with FreeBSD. OTOH, if our netstack will diverge too much
we can as well drop that idea.

--
Bye,
Waldemar

Other related posts: