Sean Ellis recently did a good presentation on this, highlighting whats different for network effect businesses:

http://startup-marketing.com/bringing-a-network-effect-busin...

Don't the fathers of lean generally believe that if you think you've got the next FB, throw away all our advice?

I believe I heard something like that on a video.

Interesting question. My take would be that you need to toe the line between building something quick and dirty, and building something that can scale. If you can't get the features right then scaling is a non issue. If you do get them right, the nature of this type of business is that it will grow quickly - whether you like it or not. So you have to be able to add scale quickly or it can fall apart due to unavailability (people often attribute friendster with this outcome). These might be the hardest businesses to start, because in addition to cracking the chicken/egg problem, you have to steer a fine line between development speed and scalability.

I think Eric's point for a business like that would be given a choice between burning time and money scaling (ie. big launch and marketing on customer acquisition) spend it on a few more iterations of refining the product, AND seeing how a small number real customers react to it.

You haven't included Digg - a perfect example of getting big instead of polishing the product.

How about Twitter? They made a product that couldn't scale, and are still doing OK.

A DB server with 256GB of RAM, 2 sockets, and multiple SSD drives can serve a lot of queries. Use that for posts, and a few slaves for reads / caching. It will cost something like $60k plus hosting fees and maintenance. But it will certainly serve 99% of web businesses as a db.

As long as your basic queries scale linearly (or maybe log-linearly), you can buy your way out of the kind of trouble most web companies would kill for.