Re: [RFC] Username/Password NSE library

["Patrick Donnelly" <batrick.donnelly () gmail com>]

On Wed, Jun 25, 2008 at 10:23 AM, Kris Katterjohn <katterjohn () gmail com> wrote:
> Patrick Donnelly wrote:
> > On Wed, Jun 25, 2008 at 12:25 AM, Kris Katterjohn <katterjohn () gmail com> wrote:
> > > My initial reactions:
> > >
> > > * I mentioned the functionality of returning a table of usernames and
> > > passwords, and the option of returning a table of only a certain size,
> >
> > Like I said, you can modify usernames and passwords to return smaller
> > tables based on a parameter (e.g. unpwdb.usernames("short")).
> > Alternatively, you can modify the design to operate on an array (and
> > use ipairs()). The user could then break on the number he wants e.g.:
> >
> > for i, user in unpwdb.usernames() do
> >   if i > 20 then break; end
> >   -- do stuff
> > end
>
> I understand this, I was merely pointing out that the closure was the method
> of choice.  I mentioned it initially[1], and nobody seemed to think that
> returning a table was better[2][3].
>
> But, like I said, if returning a table is now the desired functionality, then
> your changes seem fine.

It doesn't really return the table with the intent of sharing it with
the caller, just so the for loop can iterate over it. The main
advantage to this is the memory reaches a steady state; no new
closures or tables are made in a call to the library.

> I have run into one problem which I didn't test last night: giving the library
> a non-existent filename.  Rather than returning false and giving the caller an
> error message like mine did, or returning the error message like I assume
> yours is supposed to do[4], I get this rather nasty looking error message:
>
> SCRIPT ENGINE: error while initializing script rules:
> ./nselib/unpwdb.lua:29: bad argument #1 to 'lines' (none: No such file or
> directory)
> stack traceback:
>        [C]: in function 'lines'
>        ./nselib/unpwdb.lua:29: in main chunk
>        [C]: in function 'require'
>        ./scripts/unpwdbtest.nse:7: in main chunk
>        [C]: ?
>        [C]: ?

This was just another design choice I added in. I don't feel the
try/catch mechanism is suitable for something like this. The act of
requiring the module (ex)(im)plicitly states a level of reliance. If
the module cannot get a database it can't do anything useful and
should therefore fail immediately.

I also would suggest moving the module() call after loading the
databases (where the module can error). If we allow the module to
error, we do not want the call made before any errors. module()
changes package.loaded to the library's table and will cause
subsequent calls to require (by other threads) to return an empty
library table.

Hope that helps,

-- 
-Patrick Donnelly

"One of the lessons of history is that nothing is often a good thing
to do and always a clever thing to say."

-Will Durant

_______________________________________________
Sent through the nmap-dev mailing list
http://cgi.insecure.org/mailman/listinfo/nmap-dev
Archived at http://SecLists.Org