Re: [JDEV] CLI-P

[Jeremie <jeremie@jabber.org>]

There is one more option similiar to your #1 that I've been working
through for the JNIX proposal:

A local background server listens on a configured port for incoming
localhost connections.  This background server simply performs connection
caching per user, with a little overhead for maintaining status and
buffering data.  Whenever a command line app connects it must send the
normal authentication packet before it can reuse it's connection, so the
authentication packet is the key.

I have a proposal mostly written up, and there are quite a few issues
here, enough to make this an entire project in itself(hence: JNIX or
Jabber+Unix).  I'll create a team and set up a list to discuss these
issues next week.

Jer

On Thu, 15 Jul 1999, Jerrad Pierce wrote:

> so which provide an acceptable level of security? etc.
> 
> 1)
> a process authenticates itself to the jabber server
> and then creates a named pipe somewhere in the user's home space (~/.jabber/)
> (~/.jabber is mode 700 and the pipe 400)
> 
> the message sender talks to the authenticated process through the named pipe
> the authenticated process only sends the process if the uid of the sender
> matches the uid of the user that spawned the authenticated process...
> 
> 2)
> the authentication process gets run in the background, on recepit of SIG
> it forks a child, the child receives input and returns it and exits...
> (since the parent is in the bg, the child will be too)
> 
> advantage: the code only get's compiled the initial time it is run...
> problem: currently to send a message you'd have to:
> 
> killall -SIG (authenticated process) [we have a shell script or alias do this]
> fg (unless someone knows a way to do this programatically, for the child only)
> provide the data for the message
> terminate the message
> ^Z (because the parent comes into the foreground along with the client
> when you fg)
> 
> when what we want is:
> killall -SIG (authenticated process) [we have a shell script or alias do this]
> provide the data for the message
> terminate the message
> 
> 3)
> we become the shell, pass along all unrecognized commands to a system() or ``
> not exactly elegant...
> 
> jp
> --
> Please take the time to answer the simple survey at:
> 					http://www.pthbb.org/cgi.bin/survey/
>                                                   *             __    *      .
>            \     |     /           .        . .           .  . ((_
>                _____                       .                         .     .
>           --  /     \  --           .        .      .   +    .   .     _/\
>       oooooooooo.   |               * .       .   .           *       / ;M\_ .
>    .oooooooooooo.oo.	              .  .    .       . /\    .      /  :IMM\
>   ..oooooooooooo..oo.	Jerrad Pierce              /\  /  \         /   ;IIWMM
>   ..oooooooooo.......	209 North Street     +    /  \ /  \  .     /   ;IIIIWM
>   ...ooooooooo.......	Randolph, MA 02368        /  \     \  ___/   :;IIIIIWM
>   ....ooo....o.......	                         /    \    \ /  ::     ;;IIIMI
>    .....ooo......... 	http://www.pthbb.org     /    \     \   :     :::;IIIM
>      ..ooooooo....	__________________________ ||   ||       ::.....::::::
> MOTD on Sweetmorn, the 50th of Confusion, in the YOLD 3165. Celebrate Confuflux!:
> 
> "Here's what I know," said the physicist. "If we lived in a microscopic world, trucks would crash into walls, fly apart and then reassemble perfectly on the other side of the wall. This has been proven."
>