so I did STATISTICS! to it

[solarbird]

I did some stats tonight on my last few band blog posts, the ones that make up most of my typing output these days, and get echoed here? A couple of really interesting things popped out.

1. IPv4 addresses are no longer useful for identifying uniques or non-uniques. Almost completely orthogonal at this point, as ISPs squeeze more and more routing through fewer virtualised IPs. I have vast swathes of obvious and clear uniques coming from single IPs.

2. NOBODY, and I mean NOBODY, clicks out of Facebook. In my last two weeks, I have one Facebook-sourced load of a non-preview picture. One. Twitter's not much better, but I didn't expect it to be a traffic source. I need to completely rethink how I (try to) use Facebook.

3. I still have a lot of RSS users! Yay!

4. HI TUMBLR! I DIDN'T KNOW YOU CARED BUT I LOVE YOUUUUUUUU! <3

5. Dreamwidth matters! Yay!

6. Livejournal matters a lot. Still. HI GUYS! I'm surprised too! A regular subset of people are popping in at journal top level, but most is friendslists and hey, did you know friendsfriends still gets used? Surprise!

Counting only non-bot pageviews which view the post closely enough that non-headline images are loaded, a popular post will have well over 400 unique viewers - Montréal et Racines pulled in 437. A less popular post - something technical, like the DIY series - will pull in less, around 250. The average is a bit over 300.

What percentage of people are actually reading, I can't say.

I really have no idea what Facebook is for at this point. I need to write that "how Facebook ruins everything" post. Maybe next week.

Comments

[angelwolfgeek]

To make things more confusing, some ISPs, like mine have proper internet addressable IP addresses (not nat) but run it through a transparent proxy like mine does. Its a royal pain in the ass.

Gawd, I wish we'd just switch over to IPV6, and throw each user a nice chunk of addresses, so we could use the internet the way it was designed to be used. NAT needs to die, outside specific uses (which could be done with proper, non training pants firewalls, which modern embedded systems can handle anyhow)

[solarbird]

I predate NAT (and for that matter, we only surrendered our ancien régime Class C last year) - but I don't see a big problem with NAT, really. I mean, yes, IPV6 is totally overdue and the network infrastructure of the future? But it's been that way a while and may be for a while longer, at this rate.

I have been told by someone who is seriously good at computer security that the IPv6 stacks are first-order security clusterfucks waiting to happen. I suspect that's part of the problem. All those things what got fixed in IPv4, in v6? ... not so much.

But that's hearsay. We'll see. (I have to admit, I really miss some of those old routing methods, I really do, but, um, yeah. lolhax.)

[angelwolfgeek]

Nothing wrong with NAT. Its just a hammer, and every problem looks like a nail. You can do all the lovely security things NAT can do with a well configured router - NATs primarily are 'better security' cause they deny ports by default (but you can do the same as flexibly with public IP addresses and a properly set up firewall apparently - and you arn't forced to decide which boxes get which ports). There's a half dozen ways to get around that - my IPV6 connectivity to my dinky little toy server simply punches through my router, cause setting it properly is a pain in the rear (*grumble grumble bad documentaion grumble*), as do a lot of protocols.

ISPs natting large swathes of users to save on IPV4 addresses? Lazy. Home users using NAT to connect home networks might work for me, but I'd trade that for a good IPV6 router with a sensible interface for 90% of cases.

And then there's the old PCI DSS specifications that used to insist systems that had credit card information be behind a NAT.. they eventually switched that to network isolation.