[Netnice-developer] Re: netnice as a loadbale kernel module

[Takashi O. <ta...@wi...>]

Hi Igor,

Thanks for your interest. I'm a project leader of netnice.org.

The plan to make it a kernel module is a fairly attractive approach
for us, in that it makes the system more usable to many end users
who have never tried kernel patching. This is particularly true for
MacOS X, in that we might ask software companies to ship the kernel
module with their products to control network I/O of their applications.

But, there are several concerns about this plan. First, technically
speaking, the VIF system requires tight association between the VIF
system and OS core functionality, such as struct proc and socket
management. And, most of OS implementations do not support hooks
sufficient to make the entire VIF system "loadable". For this problem,
we may need to define a subset of VIF functionality that can be
made "loadable", or, we may need to ask core developers of the
target OSes to support hooks we need (actually, one of them once
suggested me to support such hooks to make it loadable.).

Secondly, before making it loadable, we probably need to develop a
library layer, which absorbs differences between the full-fledged
VIF system, and the (possibly simplified) loadable VIF system. This
is also good for for Windows porting. As you might have noticed,
the system uses filesystem abstraction for its API, although this
is hard to realize on Windows OS. By providing an OS-independent
library layer for general purpose network control, we can make the
system more portable, and make it more attractive to many application
developers. With the vision in mind, actually, we started the
development of the library, libnetnice, several months ago. but,
because of the nature of the open-source voluntary project style,
we are not making significant progress these days, very
unfortunately and very regrettably. We would probably need to
address this issue also.

Anyway. The idea is very exciting to us, and so, I would like to
support you as best as I can. Please let us know your feeling,
idea, and suggestions. It would be even more helpful if you let
us know background of your interest, and your intended application
field.

Again. Thank you for your message, and I would be looking forward
to hearing from you again.

thanks!


-- taka

Igor Shmukler wrote:
> Hello,
> I found your project while we were looking for a way to implement a similar functionality for our needs.
> Our original idea was to extend PF with ALTQ capabilities.
> The fact that netnice allows creation of virtual interfaces and fair-share queuing is something that interests me very much.
> I would love to know if there are plans to implement netnice and VIF as loadable kernel modules. If in that that is something that you are planning to do, we could probably contribute into the development.
> We are interested in such capabilities for pretty much all operating systems including FreeBSD 5.x, Linux 2.6.x, Windows NT (XP, 2003).
> Thank you in advance,
> Igor.
>