[Mailman-Developers] Re: Looking at performance again

[Charlie Clark]

On 2003-05-15 at 18:06:45 [+0200], you wrote:
> On Thu, 2003-05-15 at 11:49, Charlie Clark wrote:
> 
> > No, it is outbound "delivery of payload" - our application is large 
> > volume outgoing with negligible incoming. Using delivery makes sense 
> > for us in the overall sense and we use multiple SMTPs to make this as 
> > efficient as possible.
> 
> What do you do about bounce processing?  What's your bounce back ratio? I 
> would think that with deliveries skewed heavily toward outbound traffic, 
> incoming bounces will be a fairly large headache, especially as lists 
> need to be locked to register bounce scores.

Without going into implementation details - MAL is better at that and it's 
his code - we check addresses a lot better as we don't even have double 
opt-in and we have a low bounce threshold. We uncluttered the management 
interface and reduced the number of members per page and provide useful 
feedback to the admin when checking addresses including a bit of Javascript 
to highlight changes or proposed deletes.

Using an RDBMS backend allows us to use CONSTRAINTS to catch duplicates and 
other invalidities. We recently did the import and caught nearly 2.000 from 
25.000 and put them all straight on bounce. As the migration was recent and 
we don't often we don't expect to have reliable info on bounces before the 
summer. IMHO using an RDBMS for members offers lots of benefits: data 
integrity and less memory for Mailman as MAL has already explained.

Mailman has been a great basis for this work - it provides the brains for 
user management and we've been able to extend it to run SMS (text 
messaging) and other services - but we had to do a lot more than initially 
intended to achieve this. It's very much "organic" but, hey, it's Python so 
we can live with that! ;-)

Charlie