Why do some datasets need us to do `data(somedataset)` before we can use them?

Question

Why do we have to use data(spam) before we can run lda on it ? spam is a dataset in the ElemsStatLearn package.

library(ElemStatLearn)
library(MASS) # for lda
spam[5] # is ok
spam.lda = lda(spam ~ . , data = spam) # not ok
data(spam)
spam.lda = lda(spam ~ . , data = spam) # ok 

We can access spam[5] even before we run data(spam).

Answer 1

Both calls to lda work fine for me. ElemStatLearn uses lazy loading of data so you don't need to call data

> packageDescription("ElemStatLearn")[["LazyData"]]
# [1] "yes"