about a usb adsl modem driver

[Jose M Rodriguez]

El Sábado, 4 de Diciembre de 2004 01:56, Bruce M Simpson escribió:
> On Fri, Dec 03, 2004 at 09:25:56PM +0100, Jose M Rodriguez wrote:
> > But noted strong differences between atm boards and adsl modems.
> >
> > - atm boards are hardware assisted.  usb adsl modems are bare
> > assisted. - atm boards support better signaling, multiple channels,
> > ... usb adsl modems just one PVC with uni signaling.
>
> Not really. See below.
>
> > So I realize that try to get a NATM/NETGRAPH hardware driver may be
> > too work for little gain.
>
> Actually writing a NATM driver is far easier than writing a HARP
> driver. And Harti's newer drivers for ATM cards are, in fact, NATM
> drivers. You only need to implement Netgraph hooks. You don't need to
> worry about implementing Netgraph protocol nodes, generally.
>

I think so.  The actual ueagle driver do that, only needs NetGraph 
aditions.

My objections to implement NATM are related to the final operation.  
What you get througt NATM is of low interest for usb adsl users.

You don't need signaling managers, multipe channels, ...

I think that a direct, one PVC, usb adsl version of ng_atm will do the 
work.  I think implement the netgraph support directly in the usb 
driver.

About the rest of the ng_ nodes involved, I think the only problem is 
ng_atmllc.

To be a generic aal5 frame diverter, it lacks some funtionality like 
VCMux support or ppp VcMux/LLC autoswitch.

> You might want to contact Benno Rice, he was working on a software
> AAL5 kernel module to augment netnatm, which would help if you're
> targeting the lower end ADSL hardware.
>

My initial plans was implement this as part of the driver,  but I no 
have any objections to this.  But I'll prefer take this in the future, 
when we have a working driver (Well, another.  ueagle really works on 
RELENG_5).

Any pointer on this is welcome.

--
  josemi
> Regards,
> BMS