From mboxrd@z Thu Jan 1 00:00:00 1970 MIME-Version: 1.0 In-Reply-To: <56267FA6.9000502@mail.com> References: <56266870.9050204@mail.com> <56267FA6.9000502@mail.com> Date: Tue, 20 Oct 2015 11:14:09 -0700 Message-ID: From: Skip Tavakkolian To: Fans of the OS Plan 9 from Bell Labs <9fans@9fans.net> Content-Type: multipart/alternative; boundary=001a1143f84e6fae5e05228d3afa Subject: Re: [9fans] About IL Topicbox-Message-UUID: 7467bc1c-ead9-11e9-9d60-3106f5b1d025 --001a1143f84e6fae5e05228d3afa Content-Type: text/plain; charset=UTF-8 ARM and MIPS-based embedded components are cheap even for hobbyists. They usually have at least a serial interface and increasingly WiFi. On Tue, Oct 20, 2015 at 10:53 AM, Adriano Verardo wrote: > Charles Forsyth wrote: > >> >> On 20 October 2015 at 17:14, Adriano Verardo > > wrote: >> >> Could IL be actually more effective than TCP/IP in a closed net ? >> I think about a robotic application using very small cpus. >> What about Styx -- ore something similar - over IL ? >> >> >> Styx is (now) the same as 9P, and it was always similar: not a transport >> protocol, but a service protocol that ran on any suitable transport, >> and not just on IP networks. >> > Ok > >> We used a special link-level transport protocol over infra-red to use >> Styx to talk to a programmable Lego brick from Inferno. It did run-length >> encoding, and possibly some other compression scheme. >> > Possible scenarios: > 1) distributed intelligence to control complex mechanic devices. Say arms > but in general whatever else. > 2) coordination of 2+ submarine robots. Thus a very very low bandwidth > (kHz). > 3) coordination of flying drones. > >> >> All you need is a transport protocol that reliably preserves content and >> order. It doesn't need to keep record boundaries, >> although transport protocols are sometimes simpler if you do, working >> with messages instead of a raw byte stream. >> It doesn't need to be an Internet Protocol (ie, there doesn't need to be >> an IP layer). >> > Yes, I have a little experience with 9P. In a industrial appl I did years > ago, Plan9 nodes export drivers etc as a "control/monitor" file server. > The Plan9 subsystem is monitored (also) through a Windows/P9 interface. > Mission critical and a little complex but no bandwidth > constraints. > >> 9P itself will multiplex many clients >> on the same connection to a server, so you don't need a higher-level >> multiplexing protocol using ports etc. >> In fact, using attach names, you can have several different server trees >> served on the same connection to many different clients. >> > So, is it correct to say that IL is a too complex solution although > lighter than TCP/IP ? > > adriano > > > --001a1143f84e6fae5e05228d3afa Content-Type: text/html; charset=UTF-8 Content-Transfer-Encoding: quoted-printable
ARM and MIPS-based embedded components are cheap even for = hobbyists. They usually have at least a serial interface and increasingly W= iFi.


On Tue, Oct 20, 2015 at 10:53 AM, Adriano Verardo <adriano.= verardo@mail.com> wrote:
Ch= arles Forsyth wrote:

On 20 October 2015 at 17:14, Adriano Verardo <adriano.verardo@mail.com <mailto= :adriano.vera= rdo@mail.com>> wrote:

=C2=A0 =C2=A0 Could IL be actually more effective than TCP/IP in a closed n= et ?
=C2=A0 =C2=A0 I think about a robotic application using very small cpus. =C2=A0 =C2=A0 What about Styx -- ore something similar - over IL ?


Styx is (now) the same as 9P, and it was always similar: not a transport pr= otocol, but a service protocol that ran on any suitable transport,
and not just on IP networks.
 Ok
We used a special link-level transport protocol over infra-red to use Styx = to talk to a programmable Lego brick from Inferno. It did run-length encodi= ng, and possibly some other compression scheme.
 Possible scenarios:
1) distributed intelligence to control complex mechanic devices. Say arms b= ut in general whatever else.
2) coordination of 2+ submarine robots. Thus a very very low bandwidth (kHz= ).
3) coordination of flying drones.

All you need is a transport protocol that reliably preserves content and or= der. It doesn't need to keep record boundaries,
although transport protocols are sometimes simpler if you do, working with = messages instead of a raw byte stream.
It doesn't need to be an Internet Protocol (ie, there doesn't need = to be an IP layer).
 Yes, I have a little experience with 9P. In a industrial appl I did years a= go, Plan9 nodes export drivers etc as a "control/monitor" file se= rver.
The Plan9 subsystem is monitored (also) through a Windows/P9 interface. Mis= sion critical and a little complex but no bandwidth
constraints.
9P itself will multiplex many clients
on the same connection to a server, so you don't need a higher-level mu= ltiplexing protocol using ports etc.
In fact, using attach names, you can have several different server trees se= rved on the same connection to many different clients.
 So, is it correct to say that IL is a too complex solution although lighter= than TCP/IP ?

adriano



--001a1143f84e6fae5e05228d3afa--