status of misdn again

[Simon Richter]

Hi,

Tzafrir Cohen wrote:

>> chan_misdn as an external package is currently uninstallable as the 
>> asterisk config package brings a misdn.conf file.

> But chan_misdn is a separate package. It sits on top of the
> userspace misdn library. So getting it to work is a different issue.
> Right?

mISDN allows you to place the boundaries between kernel- and userspace 
arbitrarily; in TE mode, chan_mISDN uses layers 1-3 in the kernel and 
sits on top of that, while in NT mode, it uses layers 1 and 2 in the 
kernel only and uses layer 3 from the NT mode library, plus bits it 
provides itself.

> Is there actually any point in using an external chan_misdn package now
> that it has been merged into Asterisk?

Yes, the fact that mISDN's ABI changes from time to time and then it is 
better if all the ugliness of dealing with that is confined to the 
chan_misdn package.

>> The right thing to do would be hotplug integration coupled with the 
>> kernel enforcing sanity.

> Hotplug? You probably mean udev. Don't those modules load automatically
> by PCI IDs? What needs to be hotplugged?

Hotplug is the mechanism responsible for loading modules by PCI ID; udev 
just creates appropriate device nodes after the fact. I believe mISDN 
udev support works mostly (so when you load mISDN_core, the device 
appears), but hotplug support is missing (so noone loads the device 
driver), would require some rework of the driver (because the device 
driver would pull in mISDN_core only, which would then attempt to 
initialize a full stack, but has no clue about L1, L2 and L3 drivers 
yet) and probably has other issues (as with wireless LAN, we need to 
respect a country setting, for example).

    Simon