The purpose of this package is to be able to save and load only the chosen variables/columns of a data.frame in special binary files (tar archives) – which seems to be a lot faster method than loading the whole binary object (Rdata files) via load() function, or than loading columns from SQLite/MySQL databases via SQL commands.
The performance improvement gained by loading only the chosen variables in binary format can be useful in some special cases, but be sure you really need this, as non-standard file formats are used!
See this topic for details, why this package was made up for.
You can find some speed comparison of loading data (via read.spss, read.csv, load and sqldf) in the semi-finished vignette. See an image from the benchmarks:
As it can be seen, the package can complete the task a lot faster than any other data-loading mechanism in R I know. Of course no performance gain is possible in most circumstances (e.g. loading all variables from a data frame) and this package should be used in only special environments.