The mailing list for listmasters using Sympa

[Adam Bernstein <address@concealed>]

Pat --

> I've read everything I can in the manual on bounce management and it's all
> rather confusing. It would seem that there has to be a way to configure a
> list to automatically delete subscribers whose addresses are no longer valid.
>
> This is the example situation we have - a large list which is used to send
> a monthly newsletter to international researchers. So there is only one
> email per month on the list. The number of bounces will only ever be one.

Yes, we've always struggled a bit with Sympa's bounce management 
algorithms, mostly wishing like you that it was a much simpler matter of 
"if you get a single hard bounce, remove them; if you get X soft bounces, 
remove them".

One issue you're running into is the desire of Sympa to look only at the 
recent period (10 days or whatever) in evaluating bounces, which is a good 
idea except that the time period is just far too short for you, and I would 
say in general.  We've increased it on our system to 30 days; unfortunately 
this one is also not a simple parameter change in sympa.  To change it, you 
need to override the default expire_bounce._global task:

1. Create a ~sympa/etc/global_task_models directory
2. Copy ~sympa/default/global_task_models/expire_bounce.daily.task into the 
new directory
3. Edit ~sympa/etc/global_task_models/expire_bounce.daily.task, and change 
"expire_bounce (10)" to "expire_bounce (30)" (or any number of days -- you 
might want 60 or 120 or 365 or...)
4. Delete the file ~sympa/spool/task/*.expire_bounce._global
5. Restart the task-manager.pl process, to auto-recreate that task file

At least I think that's the best procedure -- I would be happy to be 
corrected if there's an easier way.

Once you've done that, the bounces will be kept around for long enough to 
start adding up to a real bounce score, and you can probably tune the 
bounce threshold parameters to get pretty close to what you want. But 
you'll never quite get there.  I too would put in a vote for radically 
simplifying the Sympa bounce evaluation algorithm, which is very smart but 
possibly too smart; sometimes dumber is better, eh?  :)

     adam