You would never test for normality that way in part because no data set is probably ever perfectly normal (unless you create one to be that).
If you are going to raise the issue of normality I would talk of one of the formal tests of normality or, better, use a qq plot to test for it. If all you are going to do is discuss the mean and median (not talk about the nature of the data) I would not even raise the point.
Just by looking at the histogram, it doesn't look like a perfect normal distribution, but for most cases, you don't need a perfect normal distribution
For example, this is perfect data for t-test, ANOVA test, or average CI
I would also try to draw a Q-Q plot and run the SW test.
Can you describe why you are interested in the distributions? Is this for an assignment, descriptive stats, or preparation for more involved stats? Also, is this real data or data for a class you are taking?
Why do you keep editing and deleting your posts, it makes it hard for us to reference and for other to learn from the thread in the future.
Many of the normality assumptions in parametric procedures are based on the errors in the models not the actual data.
If data are symmetrical enough and appear normal you go with parametric descriptives. If they aren't, you go can try to transform them into normalish data or use measures like median with interquartile range or median absolute deviation.