Hello,
Patrick Ouellette wrote:
> On Tue, Apr 27, 1999 at 03:56:49PM +0200, Jonathan NAYLOR wrote:
> > > I've watched the discussion about the need for compression to be added
> > > to the various services run over packet radio. Why is everyone so
> > > intent on modifying *every* service individually? What we need is to
> > > have the ax.25 protocol layer negotiate compression on connect.
> >
> > .... and there is the problem. If the mapping was one-to-one between AX.25
> > connections and services then we would possibly be OK. But a single AX.25
> > connection can carry many things, not just user-to-user data, for example
> > multiple NET/ROM, ROSE and TCP/IP sessions.
>
> Which is exactly the reason I think compression at the ax.25 level is the
> place to put it. Then every packet transmitted could (but does not have
> to) be compressed. We have compression available in slip, and ppp why not
> expand it to ax.25.
AX25 is not prepared to support compression. Because compression "needs"
a kind
of negotiation. Brian ko4ks and Thomas A. Moulton W2VY (Uhm, not sure it
was him)
have done a study about that. The solution was to modify a bit the ax25
protocol, using reserved bit for negotiation purposes.
Well, I think the better solution is to put an intermediate compression
layer, just on the top of ax25. That is, some of the ax25 saps (= some
of one of our callsigns+ssid) can be connected to this layer. All that
using precedent studies, sure.
On the other hand, view from the computer (higher levels), this layer
could have a similar interface than "pure" ax25, so just a new kind of
socket could be created for that, with exactly the same behaviour than a
"pure" ax25 socket.
Or even better, the compression layer could be called by the kernel when
the client uses some predefined ax25 callsigns (callsigns that we want
to use to connect to ax25 peers using compression), for example
translating the systems calls in system calls for the compression layer.
The interesting point would be to make possible have 2 ips interfaces on
the same radio, an ip address for compressed operation, another for
non-compressed. That means that both use the same broadcast callsign
(QST-0).. Oh, another idea rises to my mind: we could define a special
broadcast callsings (QST-1, QST-2), to do possible that several ip
interfaces share the same link layer, witout interferences. Uhm, not
sure of the consecuences of that.
Pd. I never receive Jonathan Messages in the list, only the replies.
Looks strange.
--
Saludos de Juli�n
-.-
