Viraptor's dev-log

… and official stats on last.fm (or lack of them) are lame. But I like stats, because stats are fun.
20 min. of mono coding and 6 hours of profiles indexing later results came. So what’s interesting in last.fm profiles? Mostly country and sex profiles. Every country has similar number of users hiding sex, but m/f ratios are very different. Mainly:

• Finland is Winland: 29.4118% M <-> 70.5882% F
• Switzerland is a don’t-go-there-man country: 92.8571% M <-> 7.14286% F
• General proportion is about 60% M <-> 40% F

Per age: (% of M) minus (% od F):

• unknown = 20.9581
• 13: -5.26316
• 14: -13.3333
• 15: -11.5385
• 16: -4.89914
• 17: -1.96937
• 18yo: 7.89981 (google spam ;)
• 19yo: 22.7621
• 20: 17.5758
• 21: 23.8994
• 22: 32.0197
• and grows… 30: ~70

Young girls like last.fm? Per number of friends:

• 00-05: 25.2016
• 05-10: 22.0175
• 10-15: 8.92857
• 15-20: 15.5963
• 20-25: 8.49673
• 25-30: -10.1449
• 30-35: 7.54717
• 35-40: 6.52174
• 40-45: 5.61798
• 45-50: 9.52381
• >50: -23.0769

Girls have more friends. Splitting friends by country gives nothing and is more useless than this whole post. More stats will follow, if there’s more stuff found.
C# library for interacting with last.fm will be published as soon, as I change separator to something other than “,” (thanks “Taiwan, Republic of China”). Currently it can poll profile details, friends lists, feeds, shouts and aggregate that in db4o.

Based on “The Spinning Cube of Potential Doom” and many other similar projects - I present Spam Cloud. What’s Spam Cloud about? It’s just pretty visualisation of spamming areas of internet. It shows hosts as blinking points for each spam record in your log. Recipe to make Spam Cloud:

1. Get your /var/log/maillog.
2. grep all lines with “Rejected: you’re in (hostname) rbl list”
3. Get source ips.
4. Map a.b.c.d address into 4d space: a->x, b->y, c->z, d->time
5. Hack a short program in C / OpenGL and take screenshots:

Read the rest of this entry »

So… mail stats.

I’ve been using dnsbl.sorbs.net, sbl-xbl.spamhaus.org and some less known rbls. These 2 host made it to config v.2, as they stopped 99.9% of spam. Other rbls have caught at most 20 messages since july 2005. It just doesn’t make sense to keep them and wait for their queries. Anything that can be stopped, will be stopped by sorbs and spamhaus. Another thing is that they nicely add to each other. Sorbs and spamhaus stop about 75% of spam each. Together they stop almost 100%, so it’s good to keep them both.

Now some graphs. Everyone loves them :)

Read the rest of this entry »