New network implementation proposal [was: Re: ipv6 on hurd]

[Olivier Péningault]

le jeu 24-10-2002 à 09:59, Niels Möller a écrit :
> My advice is that you start by designing and implementing some kind of
> ethernetdevice (kernelspace or part kernel, part userspace translator)
> that can support running several pfinets in parallell. Perhaps it
> should look like the bsd tunnel devices, but I don't really know.
> 
> Another possible design is to have a separate ip-over-ethernet process
> that reads and writes ethernet frames on one hand, and ip-packets on
> the other (doing arp and other ethernetspecific things). But one has
> to be careful about the number of rpc calls needed to send a piece of
> data over the network, perhaps ip-over-ethernet is better implemented
> as a module in the pfinet process than as a separate process.
This is nearly the idea I had.

- A translator per physical device, wich will present the upper side of
a layer 2 protocol (ex: ethernet protocol over an ethernet card, ppp
over a serial port, ...). It will be the link between the physical
device mastered by the kernel, and upper things.
Maybe users will can replace them ?
For the moment, device drivers are in the kernel, but when they'll be in
user space, they just will have to implement one more interface.

(Then I don't want to do like in pfinet. Protocol code must not be part
of a big translator, while replacing one protocol is too hard. Here is
my idea:)
- A large set of small translators, each of them implements only the
functionalities part of one (and only one) protocol (layer 3 -> 5). They
will be attached to nodes such as:
/servers/net/protocols/(ip4|ip6|icmp|arp|x25|tcp|udp|sctp|.......)
Users will easilly can change these tanslators.
Maybe layer 2 protocols could exist here too, to build tunnels or
bridges...

- Finally, another translator (one for the whole system) named hurd-net,
that will be the center of the stack. It will be a multi-threaded,
non-replaceable, updatable by users, interface between a)users b)layer
3+ translators and c)layer 2 translators.
It will be stored in /servers/net/hurd-net

In fact, layer 2 translators will present _real_ physical devices to the
system, while net-hurd will present many interfaces (physical and
logical). Users will be able to add and use their own logical interfaces
in hurd-net.
layer 3+ translators will only be able to manipulate and transform
packets as needed, network configuration and other data will be stored
in the central translator.
To understand well how hurd-net and layer 3+ protocol interact, you can
see hurd-net as a protocol-switching and data-store translator. Real
work (except configuration data management) is done in layer 3+
translators.

To avoid the need of porting all programs in the same time if interfaces
of networking change, a translator will can be used at first to wrap
(correct word ?) pfinet interface with hurd-net interface in the
users/programs side.

Routing protocols, firewall, q.o.s. may be easy to add.

I would like people here to discuss about it, and then (but not before
two monthes) I'll begin to implement it.
Comments ? :)

olivier