Could Hotmail drop mails to save bandwidth?
Could Hotmail silently drop emails with attachments just to save bandwidth and disk space? That's what Hal Licino pretends, and it made a lot of noise.
He has an interesting case but the claim is so far fetched that we are rightfully suspicious. Since his methodology is not obvious, it's hard to reproduce the results. I decided to retry the experiment in a way that others could verify easily.
Methodology
I created 3 brand new accounts on Hotmail, Google Mail, and Yahoo Mail. I deleted the welcome messages and sent 100 emails without attachment to them. Then I counted spam and mails in the inbox. I did it again with a picture attached to every mail.
In order to have legit looking mails, I included prose from the King James Bible in the messages. The Yould distribution includes a script to download the King James Bible from Project Gutenberg. I scanned it for complete sentences then randomly picked a few ones for each of the messages.
I used this convenient script to automate the process. For the attachments, I used my pictures of the Kilauea. Those are jpeg files from 100k to 500k. I waited 45 minutes after the my local mail queue was empty before counting mails on the webmail side.
$ mkdir /tmp/kjb
$ cd /tmp/kjb
$ ~/yould/yould/dl-kjb.py
$ cat *.txt > /tmp/kjb.txt
$ cd ..
$ tar -xvf lava-pack-1.tar.bz2
$ python hammer_mail.py -n 100 kjb.txt me@example.com \
example-1@gmail.com \
example-1@hotmail.com \
example-1@yahoo.com
$ python hammer_mail.py -n 100 -a lava-pack-1 \
kjb.txt me@example.com \
example-2@gmail.com \
example-2@hotmail.com \
example-2@yahoo.com
$ sudo runq
Results
Mails without attachment:
- Google Mail: 100 mails received
- Yahoo Mail: 99 mails received
- Hotmail: 100 mails received
Mails with attachment:
- Google Mail: 100 mails received
- Yahoo Mail: 100 mails received
- Hotmail: 100 mails received
Draw your own conclusions. Don't be afraid to try the experiment yourself or to suggest an alternative methodology that would explain why Hal Licino lost so many mails.
Comments
There are a few problems with this that don't mirror the methodology of Hal's. While this could be considered a reasonable computer test, it is not a very good scientific test.
Your test doesn't at all match the usage pattern of any user, it only tests one day at one time of day, the bulk of mail it produces would not be construed as being sent manually because of the timing, and it only tries one of the four directions of mail really needed to test properly (H->H, H->O, O->H, O->O as a control).
I like that you've tried to verify it though. People should check things more often.
Amusing that Will should complain about your grammar. "Your going..."? or "You're going...?
Your going to have to fix grammatical errors before people will take you seriously.
You are painfully right. I should not publish something until I have peace of mind. I think I'll make it a policy to run Festival on every new post.