[9fans] New /prog idea

[9fans] New /prog idea

[Oleksandr Iakovliev]

[-- Attachment #1: Type: text/plain, Size: 262 bytes --]


Just idea, but seriously, why cannot do something like this:

# cat /prog/new > $id
# cat /dis/ls.dis > /prog/$id/dis
# echo "/" > /prog/$id/cwd
# echo «Running» > /prog/$id/status

Not to do it which echo/cat, but to have remote access to /prog/new

[-- Attachment #2: Type: text/html, Size: 651 bytes --]

Re: [9fans] New /prog idea

[erik quanstrom]

On Sun May  4 18:01:22 EDT 2014, yshurik@lynxline.com wrote:

> 
> Just idea, but seriously, why cannot do something like this:
> 
> # cat /prog/new > $id
> # cat /dis/ls.dis > /prog/$id/dis
> # echo "/" > /prog/$id/cwd
> # echo «Running» > /prog/$id/status
> 
> Not to do it which echo/cat, but to have remote access to /prog/new

so that's an interesting idea.  i've toyed with the idea of having
a mount driver analog for system calls, but got tripped up on
the details, and the lack of a specific need.

- erik

Re: [9fans] New /prog idea

[tlaronde]

On Sun, May 04, 2014 at 08:53:39PM -0400, erik quanstrom wrote:
> 
> On Sun May  4 18:01:22 EDT 2014, yshurik@lynxline.com wrote:
> 
> > 
> > Just idea, but seriously, why cannot do something like this:
> > 
> > # cat /prog/new > $id
> > # cat /dis/ls.dis > /prog/$id/dis
> > # echo "/" > /prog/$id/cwd
> > # echo «Running» > /prog/$id/status
> > 
> > Not to do it which echo/cat, but to have remote access to /prog/new
> 
> so that's an interesting idea.  i've toyed with the idea of having
> a mount driver analog for system calls, but got tripped up on
> the details, and the lack of a specific need.

The Bell Labs paper about plan9 talks about metaphors that can be
abused: "Nonetheless, it is possible to push the idea of file-base
computing too far." This the "Discussion" at the end of the paper...
Playing with the idea and thinking about the implementation is probably
a good way to understand the paper...
-- 
        Thierry Laronde <tlaronde +AT+ polynum +dot+ com>
                      http://www.kergis.com/
              http://www.renaissance-francaise.fr/
Key fingerprint = 0FF7 E906 FBAF FE95 FD89  250D 52B1 AE95 6006 F40C

Re: [9fans] New /prog idea

[erik quanstrom]

> On Sun, May 04, 2014 at 08:53:39PM -0400, erik quanstrom wrote:
> > 
> > On Sun May  4 18:01:22 EDT 2014, yshurik@lynxline.com wrote:
> > 
> > > 
> > > Just idea, but seriously, why cannot do something like this:
> > > 
> > > # cat /prog/new > $id
> > > # cat /dis/ls.dis > /prog/$id/dis
> > > # echo "/" > /prog/$id/cwd
> > > # echo «Running» > /prog/$id/status
> > > 
> > > Not to do it which echo/cat, but to have remote access to /prog/new
> > 
> > so that's an interesting idea.  i've toyed with the idea of having
> > a mount driver analog for system calls, but got tripped up on
> > the details, and the lack of a specific need.
> 
> The Bell Labs paper about plan9 talks about metaphors that can be
> abused: "Nonetheless, it is possible to push the idea of file-base
> computing too far." This the "Discussion" at the end of the paper...
> Playing with the idea and thinking about the implementation is probably
> a good way to understand the paper...

i think it's possible to take the papers too far.  :-)  the plan 9 group was
in my opinion one of the best ever.  but that part of the discussion is
a opinion, and was taylored for a very different world, and a different
operating system.  the opinion was about plan 9 in the early 90s.  not
inferno today when the internet is ubiquitious.  i'd be interested in the
argument that things haven't changed enough to call for some reevaulation.

additionally, i think it is very interesting to consider a plan 9-like system
that has system call namespaces.  it is perhaps a way to deal with hetrogenious
multiprocessors, or even numa "machines" with big numa differences.  like, say,
a cluster.

- erik

Re: [9fans] New /prog idea

[Charles Forsyth]

[-- Attachment #1: Type: text/plain, Size: 630 bytes --]

On 5 May 2014 13:09, erik quanstrom <quanstro@quanstro.net> wrote:

> that has system call namespaces.


the system calls that are not currently in the name space are those that do
not distribute
(and there have been a few things in the name space that don't distribute
and are troublesome).

segbrk doesn't distribute. rfork doesn't distribute. you can have a control
request in a name space that operates on another process,
or creates another process, but that's not a system call by the calling
process: it's a request to another process for it to do something
(possibly in the kernel domain, if there is one).

[-- Attachment #2: Type: text/html, Size: 1090 bytes --]

Re: [9fans] New /prog idea

[erik quanstrom]

> the system calls that are not currently in the name space are those
> that do not distribute (and there have been a few things in the name
> space that don't distribute and are troublesome).
>
> segbrk doesn't distribute.  rfork doesn't distribute.  you can have a
> control request in a name space that operates on another process, or
> creates another process, but that's not a system call by the calling
> process: it's a request to another process for it to do something
> (possibly in the kernel domain, if there is one).

yes.  to implement this idea, one may need to change the set of system
calls.  or maybe like file operations, not all system calls are valid in all
contexts.  i never ment to imply this was more than a half-baked idea,
for a system that's no longer plan 9.  (if i could see how things worked,
i'd implement a toy system.)  i think the proliferation of system calls in
nemo's kernels with specialised cores might have been a related problem.

- erik

Re: [9fans] New /prog idea

[Oleksandr Iakovliev]

[-- Attachment #1: Type: text/plain, Size: 106 bytes --]

On 05/05/2014 03:06 PM, erik quanstrom wrote:
> nemo's kernels

What is the nemo kernel? by the way


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 349 bytes --]

Re: [9fans] New /prog idea

[erik quanstrom]

On Mon May  5 10:40:18 EDT 2014, yshurik@lynxline.com wrote:

> On 05/05/2014 03:06 PM, erik quanstrom wrote:
> > nemo's kernels
>
> What is the nemo kernel? by the way

i used that as a term of art to avoid the fact i messed up by not changing
the name of nix, which in 9atom is quite different, not having the ?C cores.

http://lsub.org/ls/nix.html

- erik

Re: [9fans] New /prog idea

[Gorka Guardiola]

[-- Attachment #1: Type: text/plain, Size: 830 bytes --]

There was a proc box in the original Plan B of which there were different
version in the descendants (for files too) and
which permitted things similar to what you want. See

http://lsub.org/ls/export/man.1e.ps (page 31)
and
http://lsub.org/ls/export/ubiterm.icps05.pdf

Later, Andrey and Ron did something similar for clusters with xcpu:

http://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=04100349

G.



On Sun, May 4, 2014 at 11:59 PM, Oleksandr Iakovliev
<yshurik@lynxline.com>wrote:

>
> Just idea, but seriously, why cannot do something like this:
>
> # cat /prog/new > $id
> # cat /dis/ls.dis > /prog/$id/dis
> # echo "/" > /prog/$id/cwd
> # echo «Running» > /prog/$id/status
>
> Not to do it which echo/cat, but to have remote access to /prog/new
>



-- 
- curiosity sKilled the cat

[-- Attachment #2: Type: text/html, Size: 1678 bytes --]

Re: [9fans] New /prog idea

[lucio]

[-- Attachment #1: Type: text/plain, Size: 360 bytes --]

It seems to me that this type of operation would be considerably more
successful if there was better interaction between the kernel and a
running program.  that is, if processes were engineered to accept
unsolicited signalling from the kernel and, consequently, from other
processes without being restricted to a narrow range of signals or
notes.

++L

[-- Attachment #2: Type: message/rfc822, Size: 7081 bytes --]

[-- Attachment #2.1.1: Type: text/plain, Size: 830 bytes --]

There was a proc box in the original Plan B of which there were different
version in the descendants (for files too) and
which permitted things similar to what you want. See

http://lsub.org/ls/export/man.1e.ps (page 31)
and
http://lsub.org/ls/export/ubiterm.icps05.pdf

Later, Andrey and Ron did something similar for clusters with xcpu:

http://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=04100349

G.



On Sun, May 4, 2014 at 11:59 PM, Oleksandr Iakovliev
<yshurik@lynxline.com>wrote:

>
> Just idea, but seriously, why cannot do something like this:
>
> # cat /prog/new > $id
> # cat /dis/ls.dis > /prog/$id/dis
> # echo "/" > /prog/$id/cwd
> # echo «Running» > /prog/$id/status
>
> Not to do it which echo/cat, but to have remote access to /prog/new
>



-- 
- curiosity sKilled the cat

[-- Attachment #2.1.2: Type: text/html, Size: 1678 bytes --]