I hope I will ot repaet the other answers too much.
When using Lr, you have 2 complete different sets of data. The first opne is the raw file and the second one is / are the steps you did when you developed it. As you can image, the first dataset is a constant which doesn't chnage ever, the second changes everytime you do any developments.
The finished developed ( resultng) image can be seen in Lr as well as printed without the need of creating any physical file like jpg or tif. When you print a file in Lr, Lr reads the raw, applies the changes from dataset 2 and then does the printjob very well. When you want to post it to any site, you will need a physical file, and this can easyly be done by export, and, as you imagine, it consists of the raw data + the developement changes.
This concept has myn advantages - and to be honest, I personally don't see any disatvatages !
Your fimage exists only one time - and you will be able to create any output within seconds. You don't have dozends of jpg lying on your computer with different sizes and dpi for diffeent purposes, just one image file and a set of develpoement steps.
When yo decide later that the developement of a file was not as good as you want, simply go in the history to the step were you want to restart and redevelope it. Pretty simple - !
But, tehre is one important thing to consider - backup both the pictures and the database ( dataset 2 in my example) !