The Code Train

Where Neil Crosby talks about coding on the train…

RSS Entries

The Social Graph, and why rel=“me” is doomed to failure

Posted on November 24th, 2008 by Neil Crosby. Filed under Blog Posts.

There has been a lot of talk recently of The Social Graph, the amorphous network of links on the web that define who we are and how we’re connected to each other. In this post though, I’m going to mostly ignore the social aspect of The Social Graph and instead look at the personal portion of the social graph – how the various profiles which we all have across the internet link together to create one single überprofile. After all, this is the building block that needs to be put in place before you can even start to think about how people link to each other across sites.

The standard way to link the ownership of two pages to the same person is for each of them to connect to the other with a rel=”me” link. The good news here is that most social sites on the web add this attribute automatically when you add a link to your profile page. So, if you linked to your twitter account from your flickr account and vice-a-versa then a relationship-bot would be able to determine that both profiles were owned by the same person. On the face of it, this is all pretty simple stuff.

The problem comes when you realise that reciprocation is necessary in order to have confidence in the validity of a link marked up with a rel=”me” attribute. To put it another way, if I were to link to Tantek’s homepage with a link containing a rel=”me” attribute that wouldn’t mean that it should be taken on trust that I was Tantek. After all, it’s blatently obvious that I’m not.

So why is requiring reciprocation a problem?

Well, most people don’t create these reciprocating links from their sites. Lets take an average internet user – for argument’s sake we’ll call him Simon Willison. He uses the web most days, and has profiles on quite a few sites. On most of those sites he’s linked back to his own site, because he knows that it’s good to let people find him. What he hasn’t done is link back to those third party profiles from his own site. What this means is that all those nice connecting links from his external profiles are sitting there unable to be relied on because he hasn’t confirmed them with a link back.

Simon isn’t alone on this either. Pulling a few names out of the bag of people I regularly bump into, the same is true for Matt Harris, David Thompson and pretty much anyone else I can think of. Even The Littlest Webdev and Jonty Wareing, whose sites literally only link out to their profiles on other sites don’t use rel=”me” to show that they are linking to other portions of their online identity.

In fact, the only people I know who are using rel=”me” (other than me) on their own personal websites are those who are heavily entrenched in the microformats community. Drew Mclellan is using them on his site, as are Ben Ward and Frances Berriman on their sites.

Actually, that’s a lie. Whilst looking around for people using rel=”me”, I did find one other person using it – Robert Lee-Cann. A big round of applause to him.

One argument many people will have against slapping a bunch of links on their homepages out to their external profiles is that adding a whole bunch of links is going to look butt-ugly. This doesn’t have to be the case though. One option would be to use iconified links back to their profiles, like Vero does. Another option would be to simply store all these links back to the profiles on a separate page on their site. This page would be linked to their homepage using a reciprocated rel=”me” link. So, all the profile’s could still point to a person’s homepage but still be indirectly verified by the tertiary page on the user’s own site.

This is what I do with my vCard page (which is built automatically from my vCard in my Adress Book on my computer) – because connects to my vCard using rel=”me”, I can tell my social profiles to link to, and then link to my social profiles from my vCard. This forms a triangle of confirmed links and Google’s Social Graph API knows who I am across the web.

But why is this so important?

Well, without a proper graph of connections a fully portable social network will never be possible. I would love to be able to sign up to a new site tomorrow, tell it who I am on Flickr or Twitter or wherever and have it be able to go off and find out who I was everywhere else and do as much prefilling of my profile as possible based on who I was everywhere else. Even more than that though, I’d love for this mythical site to tell me which of my contacts on the 30 or so other sites I frequent were also on the one I was just signing up to.

But, if even the alpha geeks of the world are not plugging into the graph, what hope is there?

Tags: , , , ,

If you enjoyed this post, subscribe to The Code Train and read more when I write more.

8 Responses to “The Social Graph, and why rel=“me” is doomed to failure”

  1. A very fair point, and it’s a key part of a whether XFN makes it from experimental to production in terms of social mapping.

    I think your /vcard page hints at the way this will go – the triangle of pages is completely logical. I see it that people’s contact pages will link to their presence on other social networks, rather than the homepages themselves. My homepage only performs this link directly at the moment because it’s so sparse.

    That said, don’t write off the idea of an actual third party service handling this linking. Check out ClaimID — (see for a thorough example) — an OpenID provider whose OpenID page provides rel=me links to everything you tell it. Your homepage, or any page, just need link there.

    Now, with a service sitting behind those links, it can offer discover for you and help you manage them. If it checks with the Google Social Graph once a week for new inbound rel-me links to any of your other URLs, if there are any unreciprocated links, it can prompt you to confirm the link (for genuine identity proliferation) or ignore it (for false positives).

    I don’t know if ClaimID does that sort of discovery/verification, but they could.

    You’re right that the idea of everyone building and maintaining these links by hand is unlikely. But I think that a generation of services managing the connections for us, built on top of XFN, is entirely plausible.

  2. I think rel=”me” is far more common than you give it credit for. Hell, I use it on my site and I’m not even a developer or microformatista.

    I suspect adoption is low right now because a) there’s little benefit right now other than ‘doing the right thing’ and b) laziness or simple forgetfulness.

    To solve this, let’s get something built that has real value (a few theoretical examples built with the SG API don’t cut it) and let’s pressure web authoring tools to support rel=”me” properly. WordPress already does, for instance. Bedroom coders and soccer moms won’t ever care about or understand XFN, so our only hope it to make it default behaviour in the tools they use.

  3. @BenWard: A “killer app”, like ClaimID is exactly the way I see these things going in the end, unfortunately none have any traction as yet.

    Auto-discovery (and likewise easy rejection), as you say, is key. Any action the user has to take to maintain these these links between sites should be kept to an absolute minimum, or they just won’t get created. This is definitely a problem-space I’m looking forward to playing in when I get some time.

  4. @Cennydd: Congrats on using it. However, you are still very much in the minority. Like you say, the joint problems of “no benefit except that you’re Doing the Right Thing” and “laziness” account for a large part of the current low levels of adoption.

    An initial useful app I’d love to see would be a registration form that pre-filled based on who you were elsewhere. It’s not the greatest app in the world, but it’s a place to start.

  5. He, I link back on my site! Great article btw.

  6. I suspect adoption is low right now because a) there’s little benefit right now other than ‘doing the right thing’ and b) laziness or simple forgetfulness.

    I’ll add a “c)”: ignorance. This is the first I’ve heard of this property, and I do web development as part of my duties. I suspect that most web devs in the real world are, like myself, “part-timers.” That is, we make web pages as part of some larger programming or media maintenance project. Before writing off this concept, you really should give mass education a chance first.

  7. Until there is a tangible benefit for the normobs of the web, only the geekiest of us will see value in this invisible “hey, I own this!”

    The row of social network links I’ve chosen to add to my blog may be reminiscent of a dodgy set of Xmas lights, but it’s a simple way to remind people that our social circles probably connect on another network somewhere.

    Inevitably, my mom asks “what’s FriendFeed? Do you not have enough friends already?” but to the more initiated, these icons also help illustrate what circles we CHOOSE to belong to (ie. notably no MySpace in my list).

    Until it’s dead easy AND normobs see the value of signposting your other web real estate, it’ll only be us geeks, I’m afraid!

  8. Agreed that this is a problem which is why the tools need to make it easy. This is something we’ve done with TypePad so that as we add value in connecting your Flickr, Digg, Twitter, etc account with your TypePad account we expose this (if you want) on your public profile using XFN. An example would be mine at

    For people who want to use their TypePad profile (or another service like it) as their identity on the web, it solves the problem for them as they end up with bidirectional links (TP Profile links to Twitter account and Twitter homepage links to TP Profile). For others, as long as there is a bidirectional link between their website/blog/homepage and their TypePad Profile which contains possibly bidirectional links to their accounts on other services then we’re helping to build out their graph with very little effort. If my blog and profile have bidirectional links and my profile links to my Twitter account then it doesn’t matter if my Twitter account links back to my blog or to my profile.

    So you’re right, it’s not a solved problem but it is something that will only get solved as more tools – especially the social networks – make it easy.

Comments RSS

Leave a Reply

TheCodeTrain Theme by Neil Crosby, Powered by WordPress