Re: System administration tool

[Dave Stagner <dave spnz org>]

On Sun, Apr 11, 1999 at 02:52:55PM -0500, Miguel de Icaza wrote:
> 
> > * Programmers should not have to write an entire GUI to build a
> > GUI-based tool.  HTML forms are marvelous this way.  
> 
> They can use glade, the GUI designer to create their guis and provide
> code in C, C++, Ada or Perl.

And then they're writing a GUI, reinventing the wheel.  CGI-style
programming is faster, easier, and more familiar to many programmers.
Configuration tools that take user input don't need usually need to do
anything special - they need to provide text fields, list boxes, radio
buttons, and other basic widgets.  The server can send the client
simple commands and the client can reconfigure its interface on the
fly.  That's why HTML forms are the most widely used and successful
customized graphical interface tool in history, i think.  

Mine isn't the only valid approach to system configuration, of course,
and there are other models in various stages of progress (most much
farther along than mine :) The winner will be the one that gets the
most custom modules and extensions written - and i suspect that means
the one that is easiest to write new modules for.  
 
> > And i DO know basic Unix client/server
> > programming.  And of course, the fewer libraries involved, the easier
> > security will be.
> 
> You mighe end up reimplemeting a subset of CORBA for no good reason.

I'm well aware of that.  And, in the longer run, i may go for a
CORBA-based solution.  But in the short run, i don't know enough about
CORBA (especially within GNOME) to be comfortable with it.  And ORBit
still isn't mature technology.  So for now, i'll code up something
with simple unix sockets.  If i spend weeks of time bogged down
learning CORBA, it's time i won't spend on the XML files and a
rudimentary client.  A better server can be written once i prove to
myself i'm on the right track.  
 
> > Communication will take place via a unix domain socket, which
> > guarantees locality.  This reduces the complexity of authentication
> > and increases security.  Others may wish to write a TCP/IP
> > server... good for them!  The server will authenticate the client
> > independently via OS mechanisms, which may be kludges - Unix doesn't
> > have a good mechanism for authenticating over unix-domain sockets (BSD
> > 4.4 does, but i don't have that functionality in Linux or most other
> > unices).  
> 
> You can do this by having the user share a secret with the
> root-running process.  Ie, have them use a secret handshake that both
> sides know (similar to what libgnorba does).

I'll take a look at that code.  Thanks for the pointer.  :}
 
> > If you've stayed with me this long, thank you.  :}  What do you think?
> 
> Great idea.  But please consider my CORBA idea:  I want to use CORBA
> to provide an API for services in the system: not only administrative
> tools, but access to devices and services with a clean-callable
> interface instead of having a hack-per-service as we currently have in
> Unix.

I've been considering it.  Actually, it's where i started, since i
really like the underlying CORBA API in GNOME.  I stepped back from
CORBA for practical reasons on getting good work done fast, and
security worries that i don't yet know enough to resolve.  But i would
like this to eventually be a CORBA system, if possible.  

-dave

-- 
Practice beautiful randomness and act kind of senseless.