HI,

There are just a "few" small details that complicate this
dream world. Performance, security, "as if simultaniously",
performance.

On Wed, 15 Jun 2005, Dave Shield wrote:

> On Wed, 2005-06-15 at 10:30, Magnus Fromreide wrote:
> > Is there any documentation available for the north end
> > interface of Net-SNMP?
> 
> Probably not, no.
> It's not an area that we've needed to provide much explanation for.
> The number of people working at that level is relatively small,
> and we've tended to use the source :-)
> 
> There is a man page for "snmp_read" (which is at the heart of
> all this), but it's not really very informative.
> 
> > As of today the only north ends that are implemented is SNMP and
> > AgentX. The SNMP north end seems to be quite strongly entangled
> > in the library
> 
> That's fair comment.
> 
> The Net-SNMP project probably suffers somewhat from the
> incremental nature of its development.  The current status
> isn't the result of clear design - it's "just growed".
> So the "north end" code was implemented within the library
> because that was the whole purpose of the package - the
> agent received SNMP requests, which were unpacked by the
> SNMP library.
> 
> It was really only with the addition of the AgentX support
> that we needed to consider other possible mechanisms for
> receiving requests.
> 
> 
> >  and the AgentX one is not easy to follow either.
> > 
> > I dream of an agent built from modules where we could have
> > multiple north ends (SNMP, AgentX, SMUX, whatever) that
> > connect to a dispatch layer
> 
> The framework for that is more-or-less in place, and the
> AgentX code (however convoluted) is probably the best
> starting point.
> 
> If you look at 'agentx/master.c' or 'agentx/subagent.c',
> you'll see that they open the session handle using the
> call "snmp_open_ex()"  (rather than the simple "snmp_open")
> 
> This allows them to specify new routines to convert an
> incoming data stream into a PDU structure (agentx_parse),
> and encode a PDU structure into a suitable stream ready
> for sending out (agentx_build).
> 
> The same approach could be used with any other "north end"
> module.
> 
> 
> That basically assumes that the encoded stream will be read
> from or written to some form of file descriptor.  But the
> modular transport mechanism already support alternative
> approaches (see the 'snmpCallback' transport where the encoded
> stream is really only dummy data anyway).
> 
> I don't know enough about how Windows DLLs work to comment on
> how well these two mechanisms might be adapted.  But I hope that
> gives you a basic idea of how things currently fit together.
> 
> 
> Dave

Regards,
/david t. perkins



-------------------------------------------------------
SF.Net email is sponsored by: Discover Easy Linux Migration Strategies
from IBM. Find simple to follow Roadmaps, straightforward articles,
informative Webcasts and more! Get everything you need to get up to
speed, fast. http://ads.osdn.com/?ad_id=7477&alloc_id=16492&op=click
_______________________________________________
Net-snmp-coders mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/net-snmp-coders

Reply via email to