[9fans] /net different from sockets, but better?

[9fans] /net different from sockets, but better?

[Rahul Murmuria]

[-- Attachment #1: Type: text/plain, Size: 897 bytes --]

I am planning on porting the /net concept of Plan9 to Linux.

My Plan:
Use libfs[1] to write a synthetic filesystem in Linux, much like
securityfs[2], or /proc. This libfs based code will make calls to the TCP/IP
stack on the linux, and basically be an alternative to the Linux kernel
sockets. As a result I will expose networking using /net, instead of POSIX
to the applications.

My Motivation:
Glendix[3]

My Question:
I know that using /net instead of sockets is very different. But is it
better? Specially from the networking side of things, not from application
development point of view (which we have already established is simpler, by
example).

[1] http://lwn.net/Articles/57369/
[2] http://lwn.net/Articles/153366/
[3] http://glendix.org/   and   http://glendix.org/glendix_iwp9_2008.pdf
and    http://www.osnews.com/story/20588/

Regards,

--
Rahul Murmuria

[-- Attachment #2: Type: text/html, Size: 1339 bytes --]

Re: [9fans] /net different from sockets, but better?

[Eric Van Hensbergen]

I may have some existing code based on npfs that you can use as a
base.  It has the basic semantics down, but had at least one bug and
was somewhat incomplete.  I'll dig around and see if I can find it -
it may be of some use as a reference, particularly for how to deal
with the clone file.

                  -eric

On Fri, Feb 6, 2009 at 9:13 PM, Rahul Murmuria <rahul.is.also@gmail.com> wrote:
> I am planning on porting the /net concept of Plan9 to Linux.
>
> My Plan:
> Use libfs[1] to write a synthetic filesystem in Linux, much like
> securityfs[2], or /proc. This libfs based code will make calls to the TCP/IP
> stack on the linux, and basically be an alternative to the Linux kernel
> sockets. As a result I will expose networking using /net, instead of POSIX
> to the applications.
>
> My Motivation:
> Glendix[3]
>
> My Question:
> I know that using /net instead of sockets is very different. But is it
> better? Specially from the networking side of things, not from application
> development point of view (which we have already established is simpler, by
> example).
>
> [1] http://lwn.net/Articles/57369/
> [2] http://lwn.net/Articles/153366/
> [3] http://glendix.org/   and   http://glendix.org/glendix_iwp9_2008.pdf
> and    http://www.osnews.com/story/20588/
>
> Regards,
>
> --
> Rahul Murmuria
>
>

Re: [9fans] /net different from sockets, but better?

[J.R. Mauro]

Since I'm new here, hi everyone!

On Fri, Feb 6, 2009 at 10:13 PM, Rahul Murmuria <rahul.is.also@gmail.com> wrote:
> I am planning on porting the /net concept of Plan9 to Linux.
>

I've been thinking about this for a few months. If you'd like help,
let me know, I'd be more than happy to pitch in.

> My Plan:
> Use libfs[1] to write a synthetic filesystem in Linux, much like
> securityfs[2], or /proc. This libfs based code will make calls to the TCP/IP
> stack on the linux, and basically be an alternative to the Linux kernel
> sockets. As a result I will expose networking using /net, instead of POSIX
> to the applications.
>
> My Motivation:
> Glendix[3]
>
> My Question:
> I know that using /net instead of sockets is very different. But is it
> better? Specially from the networking side of things, not from application
> development point of view (which we have already established is simpler, by
> example).
>
> [1] http://lwn.net/Articles/57369/
> [2] http://lwn.net/Articles/153366/
> [3] http://glendix.org/   and   http://glendix.org/glendix_iwp9_2008.pdf
> and    http://www.osnews.com/story/20588/
>
> Regards,
>
> --
> Rahul Murmuria
>
>

Re: [9fans] /net different from sockets, but better?

[Rahul Murmuria]

Hello Eric!

On Sat, Feb 7, 2009 at 8:47 AM, Eric Van Hensbergen <ericvh@gmail.com> wrote:
>
> I may have some existing code based on npfs that you can use as a
> base.

I asked a friend about npfs. Well, you are probably using v9fs, and
you are wrapping over sockets. Not what I intend to do, but like you
said, it could help.

> It has the basic semantics down, but had at least one bug and
> was somewhat incomplete.  I'll dig around and see if I can find it -
> it may be of some use as a reference, particularly for how to deal
> with the clone file.

Sounds great!
Although I am working in kernel space, please let me know if you have
something for me.

>
>                  -eric

Regards,
--
Rahul Murmuria