Getting column vectors from a data.frame built from rbind lists

Question

Data.frames seem to keep the underlying structure of the data, which can be inconvenient at times. In particular, I have a data.frame constructed from two lists (combined using rbind). Because the underlying list structure is maintained, this means I can't (easily) pull a column vector from the data.frame and functions that are often run on column vectors don't work. Is there a way to convert the underlying structure of a data.frame to a more "predictable" format.

Here's a simple example:

df1 <- data.frame(a = 2:1, b = 4:3)           # "standard" data.frame
df2 <- data.frame(rbind(list(a = 2, b = 4),   # "sad" data.frame
                        list(a = 1, b = 3)))

A quick check of str(df1$a) and str(df2$a) show that the former is a vector and the latter a list. As such, this confounds some common things we might try to do with a data.frame, for example, trying to order the data.frame by a:

df1[order(df1$a), ]  # works fine
df2[order(df2$a), ]  # returns: Error in order(df2$a) : unimplemented
                     #          type 'list' in 'orderVector1'

I know I can work around these issues with unlist, e.g., df2[order(unlist(df2$a)), ], but I'd really like to be able to make df2 "act like" df1. Especially since it is likely, in my case, that someone else may end up using my code. Is there a simply way to accomplish this (without unlisting each column one after the other)?