A friend is a friend is a friend ...

... or at least from the point of view of LinkedIn, but I think they're typical.

It's not unusual to have dozens of links ("friends") on a social networking site, even if you're not trying that hard. If you are trying, you can easily get hundreds. Are these all close personal friends, people you'd walk through fire for if they so much as asked you to? Probably not. Some of them are going to be closer than others, but there doesn't seem to be any way to indicate that.

Should there be? On the one hand, it would be useful, when chasing through your connections, to have some idea of whether friend A's friend B is someone A knows really well, or just an old schoolmate who happened to extend an invitation and, well, you wouldn't want to just say no for no reason, would you? So why not let members assign a degree of "closeness" to any friend? The resulting graph would be richer and more informative, and not appreciably harder to handle from an algorithm-geek point of view.

But would this really help? Everyone would have to make a snap judgment about "closeness" every time they added a link, and everyone will have their own slightly different idea of how "close" (say) a "5" is. Worse, ratings will almost certainly change over time, particularly on a purely social site like MySpace or FaceBook. Keeping the "who's in/who's out" numbers up to date could turn into a major timesink, not to mention an intricate political maze (but maybe that's what a lot of people are looking for on the sites in the first place?).

On the other hand, should you be looking at your connections' connections in the first place without talking to the person in the middle? Even if you decide to do that, there are still other cues to go by. On LinkedIn, for example, you can compare people's profiles and get some idea where the intersect, and you can look for recommendations. I would expect MySpace and FaceBook have more finely-developed mechanisms and conventions, but I don't know first hand (see previous comment on timesinks).

[The friend model doesn't seem to have changed much, but language and custom have adapted -- a Facebook friend is probably not your BFF -- D.H. Dec 2018]