Re: Drawterm and Inferno? Was: [9fans] fortune-worthy

Re: Drawterm and Inferno? Was: [9fans] fortune-worthy

[Skip Tavakkolian]

>>>it not be relatively easy (or at least, easier than it was) to rewrite
>>>drawterm in limbo?
>
> i'd have thought it would be close to a shell script,
> given some commands that roger has done to interact
> with rio, that might (or might not) even be in the current distribution.
>
> much of drawterm is similar to a large subset of emu
> (and supporting libraries).

What about a VM for the simplest machine possible that can run Plan9?
Add a C compiler for the architecture and then everyhting just runs.

Re: Drawterm and Inferno? Was: [9fans] fortune-worthy

[paurea]

Skip Tavakkolian writes:
 > What about a VM for the simplest machine possible that can run Plan9?
 > Add a C compiler for the architecture and then everyhting just runs.

Vmware does that job for the moment :-).
--
                 Saludos,
                         Gorka

"Curiosity sKilled the cat"

Re: Drawterm and Inferno? Was: [9fans] fortune-worthy

[Skip Tavakkolian]

>> What about a VM for the simplest machine possible that can run Plan9?
>  > Add a C compiler for the architecture and then everyhting just runs.
>
> Vmware does that job for the moment :-).

It doesn't need everything in VMware. Something simpler.

Re: Drawterm and Inferno? Was: [9fans] fortune-worthy

[boyd, rounin]

> It doesn't need everything in VMware. Something simpler.

    Boris the Blade:  Heavy is good, heavy is reliable.  If it does not work
    you can always hit him with it.

,x/[Hh]eavy/c/simple/
,x/hit him with/c/debug/

Re: Drawterm and Inferno? Was: [9fans] fortune-worthy

[Skip Tavakkolian]

What if the VM presented some things at a higher level, not devices
but filesystems.  Leaving authentication aside, on the kernel side,
they would look like foreign filesystem imported at boot.  With well
defined interfaces one could add new filesystems to the VM by
adding a dll; not much different than what is done in drawterm
anyway.

Re: Drawterm and Inferno? Was: [9fans] fortune-worthy

[Charles Forsyth]

[-- Attachment #1: Type: text/plain, Size: 397 bytes --]

might as well use/do an emulator for an existing architecture that's
reasonably straightforward, such as MIPS (perhaps with simpler
`kernel mode' interfaces).   unfortunately, you'd still need to do
virtual devices and those would need to turn round and invoke
the underlying systems's primitives, to access graphics and files for instance,
so it's not clear that it simplifies things much.

[-- Attachment #2: Type: message/rfc822, Size: 2629 bytes --]

From: "Skip Tavakkolian" <fst@centurytel.net>
To: 9fans@cse.psu.edu
Subject: Re: Drawterm and Inferno? Was: [9fans] fortune-worthy
Date: Fri, 19 Dec 2003 22:16:58 -0800
Message-ID: <56a2faa27b020228835e32308bc0acb2@centurytel.net>

>>>it not be relatively easy (or at least, easier than it was) to rewrite
>>>drawterm in limbo?
>
> i'd have thought it would be close to a shell script,
> given some commands that roger has done to interact
> with rio, that might (or might not) even be in the current distribution.
>
> much of drawterm is similar to a large subset of emu
> (and supporting libraries).

What about a VM for the simplest machine possible that can run Plan9?
Add a C compiler for the architecture and then everyhting just runs.

Re: Drawterm and Inferno? Was: [9fans] fortune-worthy

[Russ Cox]

> What about a VM for the simplest machine possible that can run Plan9?
> Add a C compiler for the architecture and then everyhting just runs.

crawls, perhaps.

Re: Drawterm and Inferno? Was: [9fans] fortune-worthy

[Skip Tavakkolian]

> crawls, perhaps.

ha! ☺

Yes, but then the requirement to run on faster and faster machines
puts in good standing with Intel.

Re: Drawterm and Inferno? Was: [9fans] fortune-worthy

[phillip stanley-marbell]

I recently (yesterday) got most of the inferno/utils to compile
for a device (a second VM for inferno, models Hitachi SH3)
that i did a few years ago. The obvious neat thing is that these
C programs can do things like draw on the screen via /dev/draw.

This means I now have [01258k][acl] which run over inferno.
I currently build the device into the Windows, MacOS and Linux
emu binaries. Integration into a native kernel is next (I had some
issues when i originally tried a few years ago, and lost interest).
It will be nice to have a native inferno that can build itself.

There are still a few rough edges i need to clear up, and I'm obviously
doing this for reasons best understood by myself. The speed is not
that great (currently about 10 MIPS on a P4 1GHz), but there are
ways to get almost an order of magnitude speed improvement.

I have a few other unix programs already running on it:
	bzip2, gzip, mpeg2enc, gcc (actually, just cc1).

The source for the device has been on my web page for a while,
I just recently upgraded it to fit into inferno4e.

cheers,
pip


On Dec 20, 2003, at 7:40 AM, Charles Forsyth wrote:
> might as well use/do an emulator for an existing architecture that's
> reasonably straightforward, such as MIPS (perhaps with simpler
> `kernel mode' interfaces).   unfortunately, you'd still need to do
> virtual devices and those would need to turn round and invoke
> the underlying systems's primitives, to access graphics and files for
> instance,
> so it's not clear that it simplifies things much.
>
> From: "Skip Tavakkolian" <fst@centurytel.net>
> Date: December 20, 2003 1:16:58 AM EST
> To: 9fans@cse.psu.edu
> Subject: Re: Drawterm and Inferno? Was: [9fans] fortune-worthy
> Reply-To: 9fans@cse.psu.edu
>
>
>>>> it not be relatively easy (or at least, easier than it was) to
>>>> rewrite
>>>> drawterm in limbo?
>>
>> i'd have thought it would be close to a shell script,
>> given some commands that roger has done to interact
>> with rio, that might (or might not) even be in the current
>> distribution.
>>
>> much of drawterm is similar to a large subset of emu
>> (and supporting libraries).
>
> What about a VM for the simplest machine possible that can run Plan9?
> Add a C compiler for the architecture and then everyhting just runs.
>

Re: [9fans] fortune-worthy

[David Presotto]

[-- Attachment #1: Type: text/plain, Size: 856 bytes --]

>
> (Hard to believe that the birthplace of Unix will be forced by management
> to use Windows!)

Research won't be forced to live on it.  However we already are forced to
use it to communicate with those who don't really care what the solution
is, which is most of the company.  Also, if we want to solve peoples'
problems within the company, we have to do it in their context even if
our initial experiments are on Plan 9.  To boldly go, we can use Plan 9.
If we want anyone in our corporate world to follow, we have (or
someone else has) to figure out how to use our solutions in the
other contexts.

For most people, a keyboard and a manual are both fearful things to be
avoided at all cost.  Windows does a pretty good job of making an interface
that can be more or less guessed at by trial pointing and clicking and
lots of prompting.

[-- Attachment #2: Type: message/rfc822, Size: 3033 bytes --]

From: Brantley Coile <bwc@borf.com>
To: 9fans@cse.psu.edu
Subject: Re: [9fans] fortune-worthy
Date: Wed, 17 Dec 2003 09:52:50 -0500
Message-ID: <05765b085429865f54c42d315b473a94@borf.com>

	There's also the constant friction caused by not using the solution of the
	day for embedded systems.  It used to be only Vxworks.  Now its slowly opening
	up to Linux because its the flavor of the week at IBM and Lucent looks at
	IBM as the model of a reincarnated big company.  The GPL scares management
	somewhat which acts as a damping force.

A problem with using Linux in embedded systems, and this would be true of any
large source-only system, is that while it give the developers ready made tools,
it also limits their solution to those tools.  I watched one company use Linux
pretty closely and it worked, but the solution is couched very high up.
My first Embedded Unix product was shipped in 1988.  It was based on V7 and
was only 70K or so of kernel and a couple of small programs.  The kernel was
adapted to the applicaion.  Worked really well.  I recently looked into
using Linux for an embedded system, but that didn't work out.  It's not
documented.  I found it frustrating.  Dropped it and went back to my
own small kernels and just wrote more code, which I did faster than
I could figure out how to get Linux to do what I wanted.

Observation: A sufficently large amount of source == no source at all.

(Hard to believe that the birthplace of Unix will be forced by management
to use Windows!)

Brantley

Re: [9fans] fortune-worthy

[Skip Tavakkolian]

>> (Hard to believe that the birthplace of Unix will be forced by management
>> to use Windows!)
>
> Research won't be forced to live on it.  However we already are forced to
> use it to communicate with those who don't really care what the solution
> is, which is most of the company.  Also, if we want to solve peoples'
> problems within the company, we have to do it in their context even if
> our initial experiments are on Plan 9.

Isn't this an argument for stronger drawterm effort (or something
more ambitious)?

BTW, I'm assuming this is referring to Windows for desktop environments
and not for running servers. right?

Drawterm and Inferno? Was: [9fans] fortune-worthy

[Martin C.Atkins]

On Wed, 17 Dec 2003 23:57:56 -0800 "Skip Tavakkolian" <fst@centurytel.net> wrote:
> Isn't this an argument for stronger drawterm effort (or something
> more ambitious)?

This is probably a silly thought, since I haven't researched it, but...

Given the recent changes to Inferno (fileserving, graphics, etc), would
it not be relatively easy (or at least, easier than it was) to rewrite
drawterm in limbo?

The advantages are obvious - a single drawterm binary would work
identically on all the platforms supported by Inferno!

Martin
--
Martin C. Atkins			martin@parvat.com
Parvat Infotech Private Limited		http://www.parvat.com{/,/martin}

Re: Drawterm and Inferno? Was: [9fans] fortune-worthy

[Charles Forsyth]

>>it not be relatively easy (or at least, easier than it was) to rewrite
>>drawterm in limbo?

i'd have thought it would be close to a shell script,
given some commands that roger has done to interact
with rio, that might (or might not) even be in the current distribution.

much of drawterm is similar to a large subset of emu
(and supporting libraries).

Re: Drawterm and Inferno? Was: [9fans] fortune-worthy

[Martin C.Atkins]

On Fri, 19 Dec 2003 08:18:00 0000 Charles Forsyth <forsyth@caldo.demon.co.uk> wrote:
> i'd have thought it would be close to a shell script,

That would reduce the complexity somewhat - assuming it isn't a
*really* horrendous shell script :-)

> given some commands that roger has done to interact
> with rio, that might (or might not) even be in the current distribution.

Yes, I was remembering Roger's Charon-in-a-window work, and hoping
that it implied good things about the possibility of reversing the roles,
which is, I think, mostly what drawterm would involve.

> much of drawterm is similar to a large subset of emu
> (and supporting libraries).

I wish all my shower-time ideas turned out so good!

Martin
--
Martin C. Atkins			martin@parvat.com
Parvat Infotech Private Limited		http://www.parvat.com{/,/martin}

Re: Drawterm and Inferno? Was: [9fans] fortune-worthy

[boyd, rounin]

> Given the recent changes to Inferno (fileserving, graphics, etc), would
> it not be relatively easy (or at least, easier than it was) to rewrite
> drawterm in limbo?
>
> The advantages are obvious - a single drawterm binary would work
> identically on all the platforms supported by Inferno!

sounds like a plan and you could bugfix it without, say, a µ$loth SDK.