Not to discount the skills of Joel and Jeff, but one thing to keep in mind is it's much easier to build a social site for a bunch of hackers, than for example, a bunch of retirees.

The same thing goes for Hacker News. StackOverflow and HN are great, but when you look at their traffic figures, (1600 a day to start for HN, and 30k a day today) keep in mind that websites for computer programmers will just in general perform the best among any sort of website, in terms of market penetration.

What they don't perform very well in is having visitors who don't run adblock. Because of this quirk, you need to have a million-dollar revenue source like Spolsky has on his blog - a job board for programmers.

I think they missed the fact that the two people who built it are insanely famous in their target market.

Actually, they didn't - one of the 9 "building blocks" they mention was critical mass - knowing they'd have a large initial audience because of their existing blogs.

I have a hard time swallowing this. It sounds like a classic case of confirmation bias. "We did these nine things and were successful, therefore these nine things made us successful."

There's no data which tells us which of the nine building blocks mattered, if any, and to what extent.

What's more likely is that they believed these nine things were important -- and some of them quite possibly were -- and their belief has been confirmed by their success. But that doesn't mean if you duplicate these 9 features you will be successful, or that you have to duplicate all 9 in order to be successful.

All-in-all not a very helpful article.

I think they may be overestimating the value of a Microsoft stack. I think it's reasonable to get 16 million pageviews with no load on almost any popular open source stack if you are using 2 dedicated boxes each with 8-core Xeons.

If they were having almost no load on low-end virtualized hardware running a Microsoft stack, that would be noteworthy.

The article mentions their 8 core Xeon running SQL server only cost $5,000 in licensing... Either the article is wrong or they haven't appropriately licensed their SQL server. Microsoft is pretty explicit that for web-facing applications you must use per-processor licensing. Which would mean the licensing on that server should have cost $40,000 for SQL Server Standard edition.

Ugh, I shouldn't post before I've had my morning coffee.

Microsoft's licensing policy is per processor, not per core. Please disregard my statement above.

(I'm off to distribute some corrective lashings now.)

I noticed that also - even at 5k/license that is buying at least 2 more dedicated boxes to throw (name your FOSS RDBMS) on

He means ethnography, not anthropology. To notice that a bunch of backpackers sit on the steps is ethnography. To study an entire culture's way of life is anthropology.

Moreover, the argument is crap. It's a true statement at face value, but what he says is not really what he means, and his supporting observations don't actually support the statement.

This is the same guy who wrote about how design is just decoration, and drew a comparison to the flowery bits on NYC brownstones to prove it -- the architect wouldn't bother with the flowery bits, he'd pay some menial laborer to just stick shit on there. (I don't even believe that assertion, but it's a very poor supportive argument.)

Sigh.

I think having two incredibly popular cult-like blogs probably had much more to do with it than the "9 building blocks" or whatever that every other site on the internet now has.

Actually, that was one of the 9 building blocks.

Yeah, they should have just called it "marketing", but I can understand why they didn't, because their audience hates words that non-technical people like to use.