[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: n-hurd networking

From: Thomas Bushnell, BSG
Subject: Re: n-hurd networking
Date: 07 Aug 2002 12:22:02 -0700
User-agent: Gnus/5.09 (Gnus v5.9.0) Emacs/21.2

Hisham Kotry <etsh_cucu@yahoo.com> writes:

> What's are the options of providing a n-hurd with
> networking support?

My term for this is a "collective".  

> The only possible scenario I can see is to get all
> n-hurds to use the same network stack. Otherwise, some
> un-expected actions will happen that could break
> connectivity for other n-hurds, for exmaple, if every
> n-hurd starts its own hurd-net translator and reads
> all traffic entering a specific NIC, some stuff could
> happen, like:

You betcha!  Of course, only one network stack makes a bottleneck.  

My dreaming suggest the following:

One IP address is allocated to each individual Hurd, and an additional
IP address is allocated to the collective as a whole.  Generally, only
the latter is seen by most users.

The Hurd's use their individual IP addresses to talk to each other and
to send Mach (or whatever kernel) IPC with network transparency.

The Collective's IP address is managed something like as follows:

Each Hurd has a copy of the routing table.

Each Hurd has responsibility for some subset of the incoming network
traffic.  Each TCP connection, for example, is owned by a particular
Hurd.  Each SYN listener is owned by a particular Hurd.  Etc.  For
obvious reasons, the Collective tries to have the owner of a network
connection be the same Hurd on which is running the user land program
that has the socket open.

Listening to ICMP would generally be owned by only a single Hurd.


reply via email to

[Prev in Thread] Current Thread [Next in Thread]