[asterisk-dev] A plan for span

[Shaun Ruffell]

Oron Peled wrote:
> On Tuesday, 8 בJuly 2008, Shaun Ruffell wrote:
>> If the registration utility only registers devices that are configured
>> (and  they have to be configured anyway in some form to work with
>> asterisk), and if all devices that are dahdi compliant only registers
>> themselves on command, adding a new device wouldn't cause any problems
>> since it would never be registered until configured to do so and then
>> presumably the administrator would add it to the end of the list
>> of devices.
> 
> How the administrator is going to specify *it* when *it* has no name?
> The main problem is that "Span 5" does not identify any specific
> piece of equipment.
> 
> Do you think that uniquely naming spans and exposing this via sysfs
> as I proposed is a major complexity challenge? I don't think
> so and we will be happy to do this if needed.
> 
> 

No..not at all...I was only saying that with unique device naming via sysfs as you suggested, and making every dahdi driver dynamically registrable like your drivers are currently, and updating the registration script to specify a load order based on the unique name, many of the immediate goals can be solved without requiring any changes to how asterisk interacts with zaptel/dahdi. 

This would be relatively simple and solves the problems of:
1) breaking dialplans when new devices are added.
2) the silent renumbering channels when a device is removed or malfunctions (instead you'll get an error that a specific unique device didn't show)
3) races when devices are loaded in parallel.

With the added benefit of being able to get the devices started on sometimes long initialization processes early in the background before needing to register which should speed up overall system startup (since much of the time needed to start the wctdm24xxp and wcte12xp is spent waiting for interrupts on register reads and writes)

So I was basically agreeing with you, with the added input that supporting dynamic registration across all devices is a good / easy first step on our way to getting away from referring to spans and channels by number only.