Knowles (who works under game economist Edward Castronova) posits that there are three primary criticisms with the sub model: it incentivizes players to rush through content, it requires the steady release of expensive content, and it prevents companies from price discriminating (charging players for how much those players value the game rather than a flat rate). He then works to dismantle each criticism, using math as his primary weapon.
If you're into detailed looks at payment models, the full blog is worth a read. Hit the jump for a quick summary of Knowles' main points.
On rushing through content: The average subscription MMO player is in his or her 30s, married, and working full-time, meaning this player does not have time to rapidly burn up new content in the way sub-detractors predict.
On expensive content: Content is only expensive in relation to how much revenue it generates. In other words, the cost of content creation only matters if a game has enough subscribers to make the production costs worthwhile. Good content will be a profitable venture for studios working on a sub model.
On price discrimination: Games with subscriptions are able to price discriminate because individual players assign different values to subscription fees by playing a game for different amounts of time. A player who plays for two hours, for example, pays $7.50 per hour, while a player who plays for only one hour pays $15 per hour. The publisher and developer capture more value from the second player.
It's worth noting that this post is from the perspective of publishers and developers considering the sub model as a possible monetization method, not a discussion of which model works best for players.