In the case of b*, it matches zero times at the start of the string (ie before the first character) - which is as many as it can, because there are not more than 0 leading b's in your string.

In the case of b+, the empty match is not successful, so the regex engine bumps along, and tries to match as many b's as possible, starting from the second position (ie string index 1).

So both are greedy, but neither do a search for longest match of consecutive b's. try

$ perl -lwe '"abbabbbbc"=~/(b+)/ && print "Found: $1"'
Found: bb
[download]

It used the first position where b+ matched, and there took as many b's as possible.