early draft of a section from 2nd edition of blueprints
We all would like to think there was an abstract, perfect design that we could find and then never change. But different sizes demand different design approaches, and as our websites grow we have to change the wise choices we made earlier that are now liabilities. This is true of both information spaces and social spaces.
For example, everyone has seen the almost psychic spellchecker most search engines sport, but do you know how it works? It parses the millions and millions of queries and correlates when a query is made, then no click on results is made, then a second query with a large number of similar characters is made, then a click on a result. To do this and end up with a comprehensive dictionary of potential misspellings and corrections, you need millions of searches so you can identify the millions of ways people get things wrong and the millions of ways they get it right. Adding spellcheck to a website may seem easy, but if you don’t get high traffic, you can’t get the same range of suggestions and you’ll have to rely on what is likely to be a less effective approach (a discussion for elsewhere). There are many other types of websites that are changed and shaped depending on how much data they have and how many people are using it. Wikipedia is one.
Wikipedia is only interesting because of the huge numbers of people who use it. Exerts on every topic on earth join in in writing, editing, contributing citations… collectively creating the most complete entries on any topic. Because they have so much traffic, and because most people are nice, if the occasional idiot defaces a page it is repaired in under five minutes. And so goes the marketing speil, and many of the entries do indeed realize this promise. But some on each end of the spectrum of usage show their own set of problems.
The extremely popular entries or extremely controversial entries (often the same) can’t be left open to be edited by everyone, no matter what the Wikipedia philosophy is, because the number of people vandalizing it is too high to guarantee a useful entry at any given time. Wikipedia is forced to lock this entries against open editing.
Here we see a typical Wikipedia article, illustrating the power of collaboration. Ciphergoth, mlcome, OliAtlanson, Aastrup and many others are discussing how to make the article more accurate, and complete.
And here we see a page that gets almost no traffic. In fact, it didn’t exist until one day I started to wonder where the name (and the food) Jalapeno poppers came from. I searched everywhere, including Wikipedia, but all Icould find was a Chowhound discussion board article that thought they might be related to Chili Reneos. I posted what little she knew on Wikipedia in hopes that the miracles of five-minute-corrections would bring me the answer, and wandered off to ask the question on another discussion board.
People are so used to Wikipedia being extensive, complete and expert no one questioned this entry. Over the next ten months, a couple people did add to the entry, one restoring the tilde to jalapeno, another contributing a photo, and someone adding suspiciously marketing-esque information about John Neutizling’s invention of the Chile Relleno (unless he’s Mayan, I really really doubt it). That has been removed since this screenshot, but in the stub world updates are slow, and vandalism – especially subtle vandalism–remains up and the truth is arrived at with fewer miracles if it arrives at all.
Moreover, in the ten months since its creation, it is now the 4th result (5th if you could video best bets) in Google. 
The LATimes tried to leverage the power of wikis with their wikitorial. On June 17th 2005 they launched it, and on June 19th they took it down. Users were posting obscene photos and comments at a pace that no one could manage. LATimes had the large numbers needed to create interesting content, but hadn’t learned the lessons of Wikipedia’s controversial entries. After all, if Wikipedia with its vibrant and committed community couldn’t keep George Bush under control, how could a brand new newspaper section? It still hasn’t returned, and maybe it represents a problem that can’t be solved.
When you look at examples on the web to learn from, make sure you are dealing with similar problems of scale.
see also earlier size matters post