dzohrob

Readr was a social aggregator, like Friendfeed but without the Facebook acquisition. I just sold that name a few weeks ago - we'll see what the new owner has in store for it.

Seconded. Antirez, I know you hang out here sometimes, so consider this a huge thank you for building what has become a core part of my business' infrastructure.

Except if the same user clicked through to a result, the server of the result page would receive the same referrer header w/ the DuckDuckGo query in it.

Isn't this sent by the user's browser, and not by DDG? It's a client-side configuration issue -- turn off referrers in your browser. Your browser would send the same header if you clicked through a DDG search result.

Using the human body isn't a profession either. But it's likely that you and I are significantly more ignorant than a doctor regarding its inner workings despite using our bodies every day.

most people don't.

If the problem doesn't exist, why do so many people find them useful?

who wants to spend time learning to play this?

http://fretwar.com/static/data/round/1/media/xzibitonics_C.m...

Evenflow owns dropbox.com now. Check http://whois.sc/dropbox.com .

don't forget The Advantage, who do indie-rock versions of Nintendo tunes w/ live drums, bass, and dual guitars -- http://www.myspace.com/theadvantage1

You value getting it first more than getting it right. That might be aligned with TC's business, but not with the better interest of your readers or the public at large.

I'm not sure exactly what you mean by a "large degree," but the story of Detroit is more about race than it is about any of those things. (It's also about the auto industry, of course, and poor government.. but those are intertwined with racial issues as well.)

i tried signing up a few times but couldn't make it work. username too short, then the invite code disappeared on the retry page. i tried going back to step 1 and it still wouldn't let me sign up.

Read the fine print -- it's top sites by pageviews, which is a very different (and less meaningful) metric than top sites by unique visitors.

That same Alexa data you quote has POF at 168 in the US and 466 worldwide, with 21+ average pageviews/visit (Compete has them closer to 30). According to that same data, Flickr has ~9 PVs/visit, and CNN has ~3.5.

It's not inconceivable that with a few differences in the data between Hitwise and Alexa, you could come up with POF in the top 20. It just shows you how meaningless pageviews are as a metric of success or engagement.

You've obviously never been there - see http://www.forgottendetroit.com/

My only source is my business partner, but it's a pretty good test. At 10-15 feet, you can only see features as geometric shapes, but on very well-designed sites you can probably tell what you're "supposed" to do. This is most effective for signup and purchase flows, where simplicity matters most, but I think it can lead to insights on pretty much anything.

while i appreciate this from an aesthetic perspective, this fails multiple usability tests:

1) "across the room" -- stand 10 feet away from your machine and see if you can tell what you're supposed to do

2) "don't make me think" -- i'm sure anyone who has ever used a website has to think about this one.

if you actually want people to sign up, make it easier, not harder. save your creativity for your content.

what's even sadder is that the uncov kids can't handle criticism.

when i clicked through this morning, there was a user-submitted, reasonably well-written rant against the uncov guys, along with some back-and-forth in the comments. it's gone now.

don't forget the new yorker. the best $1 i spend a week. (for reference, the sunday NYT is $6 with home delivery.)

Actual traffic data, from http://www.quantcast.com/digg.com:

156.4M monthly US pageviews 224.1M monthly global pageviews

[insert standard Alexa disclaimer here]

http://www.alexa.com/data/details/traffic_details/cuil.com?s...

from the post:

"When designing UX think about what your mom would use, not what Louis Gray would use"

That's probably a good rule. That's not saying build something a geek wouldn't use.

shame on you if that's an affiliate link.

FWIW, the first one I clicked on was this:

http://www.w3counter.com/globalstats.php?date=2008-05-31

Not that it's the best source of data, but.

Many sources [1] show that IE6 still has more global usage than IE7. Though I'm sure 37signals' userbase is not represented well by global data, turning away users of perhaps the most popular version of the most popular browser is an interesting move.

[1] http://en.wikipedia.org/wiki/Usage_share_of_web_browsers

Most of the methods people use for sharding/partitioning end up in the application layer, because open-source DBs just aren't fancy enough yet to handle this.

Mysql-proxy (http://forge.mysql.com/wiki/MySQL_Proxy) looks promising, but it's still in an alpha state. You could try writing a proxy script in Lua -- there's already some master /slave stuff done in this area.

Mysql 5.1 supports partitioning, but not really in the way that you and I mean.

The problem is that sharding breaks some fundamental SQL axioms -- for example, doing SELECTs no longer returns all rows (depending on how the table is partitioned).

Almost everyone I know who does this rolls their own layer. As subwindow mentioned, if you're using Rails, you have to futz with establish_connection due to ActiveRecord's assumptions about one DB connection per class. It's possible -- we've done it -- but it requires lots of connections to the DB and it can be messy. If you're on Rails, it might be easier to scale vertically first before going horizontal.

Replication doesn't need to change in this context -- the Mysql documentation should be all you need to get started.

Start here: http://postmaster.msn.com.

I would definitely like to hear about this as well; right now we're outsourcing our payroll using PayCycle (http://paycycle.com). We're bringing on our first employee soon and definitely want to focus on our core business (ie, making cool stuff) and not administrative tasks.