1 |
Grant wrote: |
2 |
> Any drawbacks to that? Is this what you mean: |
3 |
> |
4 |
> # --lookup-by-subnet strip the last 8 bits from IP addresses (default) |
5 |
|
6 |
Yep this one and no drawbacks I can think of. |
7 |
|
8 |
> Why wouldn't the email be returned to the sender in case 1? |
9 |
|
10 |
Because number 1 is entirely composed of newsletters, automated |
11 |
responses, etc. Someone at Amazon sat down one day and realized that |
12 |
their mail queues were full of crap email. So rather than have a twenty |
13 |
server farm to send email, they took a short cut. The email is generated |
14 |
on the fly and piped directly to the socket for immediately delivery. |
15 |
I'm totally guessing here, but that's roughly how I'd do it. If the mail |
16 |
fails, the entire thing is dropped on the floor and maybe a db gets |
17 |
updated to reflect that it was never sent. You can send a ton of email |
18 |
this way because you never take the I/O hit of running the email through |
19 |
outbound MTA queues on the originating side. Because it never enters a |
20 |
"real" MTA it never gets retried when it fails. |
21 |
|
22 |
kashani |
23 |
-- |
24 |
gentoo-user@g.o mailing list |