Hi Karl,

Sorry for getting back to you so late, but at the moment i'm quite busy (also
on the pmacct side of the things: wrapping up a "stable" 0.11.6 release and in
parallel working to a couple of ideas which can give light to the 0.12 branch).
Let me summarize a little bit: 

* the ssh/rsh method is there, it just works but is inefficient;
* the socat approach is fine with me if it works: i can't advice because i
  don't know the socat package - and have no time to look into it currently.
  As said, we can make the external dependancy optional and configurable at
  compile time; if you produce some code that works at least for you, we might
  seek at improving it at a later stage for inclusion in mainstream. Or the
  patch can be made available through the CVS separatedly if it's not generic
  enough. 
* the socket approach would be, IHMO, the best but requires more work. This
  might be done in alternative or in addition to listening on a UNIX socket.
  The way in which it should be coded is similar; as operations in the IMT
  plugin are very quick (and to avoid concurrency overhead) i would advice
  something that serves only one client per time; it currently works this way
  for the UNIX socket. The reward is that once the connection is served, the
  existing code on both the client and the server can be fully re-used.
  Optionally, multi-threading can be introduced: a small but useful thread
  pool management library is already existing inside the pmacct package and
  is written by Francois Deppierraz.

One more thing: what about authentication? With an UNIX socket this is easily
solved by the underlying filesystem. Anything network-based, should require
you to take care about it (once again, can't speak for how it's going to work
for the socat approach).

Cheers,
Paolo


On Sat, Feb 28, 2009 at 07:49:50AM -0600, Karl O. Pinc wrote:
>
> On 02/25/2009 04:46:46 PM, Paolo Lucente wrote:
>
>> All this said, let me just shoot a proposal: as the in-memory table
>> client/server communication is already based on request/reply headers,
>> operation codes, etc. (so let's say the protocol is there) what about
>> keeping it simple by encoding all of this in either UDP or TCP packets
>> with a bit of socket programming?
>
> Bah.  All the time I'm spending messing around in shell makes
> me think I'd be better off extending pmacct.  I guess my reluctance
> is that if you don't like it all the work would be wasted.
> We would really have to agree on an approach.  Naturally you
> could still reject whatever code I write if you think
> it's broken or ugly.
>
> What do you think of the ideas so far?  (I do need
> to get something working quickly.)
>
> Karl <[email protected]>
> Free Software:  "You don't pay back, you pay forward."
>                  -- Robert A. Heinlein

_______________________________________________
pmacct-discussion mailing list
http://www.pmacct.net/#mailinglists

Reply via email to