Saturday, June 21, 2014

The disappearing (and reappearing) profile photo

Recently I noticed that my profile photo was broken (I've since fixed it).  "That's odd", I thought, "I uploaded it quite a while ago.  Maybe there's some glitch in Blogger's servers."  I kept checking, figuring it should come back before too long, but it didn't.  So I went to my Blogger profile to see what had happened to the image, and discovered that the URL I had given was broken.

I don't remember why I'd given a URL instead of just uploading an image.  Maybe I didn't have a copy of the image handy.  Maybe I just thought it was "webbier" to give a URL, but never mind.  Easily fixed.  I hunted up another copy of the image and uploaded it ... and we're back!

What's interesting, though, was that the URL pointed at Technorati, whose probably-no-more-tangled-than-usual history I've touched on before.  So I checked.  Technorati is still a thing, albeit clearly not one I personally pay much attention to.  Evidently they've redone their infrastructure a bit, or perhaps just cleaned out inactive accounts, causing the link to finally rot after however many years it's been since I first put it on my profile.

Links rot.  That's just part of the web.  In fact, it's a key architectural decision behind the web (as opposed to, say, Xanadu).  It would be interesting, though, to study which links rot, and when, and why.

In the case of my profile photo, a link to an obscure corner of Technorati, linked to a completely inactive account associated with a little-read blog, remained stable for years until, one day, it disappeared.  This is probably not too uncommon, but nonetheless I'd expect link rot to become less common over time.

In the old days, people would put up web sites on their personal computers, or on the workstation in their lab, and so forth.  They would get tired of the hassle of hosting the site, or graduate, or whatever, and the site would go away.  That's largely been replaced by web hosting services, but even then sites go away all the time as people get tired of paying for them and maintaining them.

However, a larger and larger portion of content is now being hosted by companies like Facebook, YouTube, Google, Twitter and so forth, or by major media outlets, which at least implicitly promise to maintain the content in perpetuity.  "Perpetuity" is rather better defined in theory than in practice, but I have a high degree of confidence that links to this blog will still work ten years from now, probably twenty and quite possibly fifty.

Will someone living a thousand years from now be able to read Field Notes?  I have no idea.  The odds of Google (or any of the other present-day giants) still being around in a thousand years are fairly small, but the likelihood of it costing peanuts to maintain everything that's ever been published on Blogger are pretty good, so who knows?

What does seem likely is that the bulk of "user-created content" will remain accessible as long as there is a web more or less like the present one for it to be part of.  If that's right, then the main sources of link rot will probably be companies folding and taking their sites down, or content owners deciding to take older content down or hide it behind paywalls or similar actions.  In other words, links are probably less likely to rot due to inattention or Life Happening to the particular person who created them in the first place, and more likely to happen due to explicit decisions by corporate entities.

Return of the cow clicker

I've previously written about Cow Clicker, a Facebook game in which players would click on an image of a cow, and later just the space where a cow had once been, thereby sending a message to all their friends that they had done so.  While not a runaway hit, Cow Clicker did manage to attract some 50,000 users, some portion of whom paid real money for the privilege of clicking more often, or on a fancier cow (Bling Cow could be yours for only $100).

The idea behind Cow Clicker was to reduce social gaming to its barest elements, partly as parody and partly as a study of social gaming behavior.  Fast forward a few years, and someone has done the same thing for mobile phone apps.  The Yo app will send a message to any of your contacts saying, simply "yo".  Unlike Cow Clicker, Yo has attracted hundreds of thousands of users so far, who have already sent millions of yos.

This popularity has had two not-too-shocking consequences.  On the one hand, it has attracted $1 million in funding.  On the other hand, it has been hacked.

Actually, the hack doesn't seem so much a hack as a matter of the app leaking confidential information and someone noticing it.  Three college students using the app were able to get the personal phone number of the founder, text him and get a call back.

What does it all mean?  Anyone who thinks it means the end of civilization as we know it is forgetting that civilization as we know it produced the tulip mania, phone booth packing, pet rocks and any number of other major and minor follies.  Nor can it possibly be surprising that an app, however trivial, that could gather hundreds of thousands of users in short order might attract investment money.  Whether or not you believe that the attention economy is anything new or different, getting people's attention is potentially worth money ... "This Yo brought to you by Spümcø".

Somewhat more concerning, though still not surprising, is that even a simple app like Yo would leak confidential information.  Security in applications of all kinds is still something you have to build in, or at least you can't assume that your app is secure just because you haven't done anything to make it insecure.  To some extent this is a hard problem.  Any useful app will involve some form of communication, and any communication exposes information, even if it's only who's communicating with whom (which can reveal much more than you might think).

It's been a couple of years since Cow Clicker's heyday.  Most likely the ruckus about Yo will die down and in another few years another minimal app will take its place.  Sic transit gloria mundi.

Thursday, June 19, 2014

The internet of loos

Auntie Beeb reports that the loos at Heathrow Terminal 2 are being fitted with sensors to detect how many people are using particular toilets, and when.

Feel free to snicker or chortle right about now.

OK, so what does this mean?  The overly harsh take would be "Yeah, that's about all this whole 'internet of things' things is going to amount to."  A more optimistic take would be "Heathrow is one of the world's busiest airports.  If they see a benefit to this, there must be something to it."  While I've seen any number of "there must be something to it" endorsements fail to pan out -- too much of this is a good sign of an impending bubble -- I tend to lean toward the second opinion.

Yes, I'm not thrilled with the term "Internet of Things", but I think that this is more because what we're seeing is a gradual trend of (some) ordinary things being put on the internet, and not a brand new phase or some sort of new internet.  Lots of things have been on the internet, some for longer than others.  Weather sensors.  Webcams.  Taxi cabs.  Temperature and voltage sensors for computers in datacenters.  As time goes on, the portion of internet data generated via human intervention will probably decrease, and the amount generated by various ... things ... will probably increase.

This isn't the hardcore IoT vision, though.  All the examples I gave are things that naturally actively generate data.  Even Taxi Cabs have always needed to communicate their location and status.  Fitting them with GPS and putting them on the net just makes that process more accurate and efficient.

The full IoT vision involves tagging everything with some sort of net-friendly identifying device, say an RFID, which can then be scanned.  If every book on your bookcase, every fork in your silverware drawer, every pair of pants in your closet and so on is tagged, then you just need to wave a scanner around in order to upload an exact inventory.

Perhaps more realistically, if newly manufactured objects carry RFIDs -- and some do -- then gradually people will come to have more and more net-visible things around them.  What we choose to do with that data is another matter, as are a number of privacy concerns (what's to keep someone from walking by your house with a scanner and seeing what's in it?).

In that sense, the Heathrow loos are more like weather sensors and taxi cabs and less in line with the "tag ALL the things" concept.  Interesting though they may be, they don't say much one way or the other about how the larger IoT vision will play out.

Sunday, June 1, 2014

And the winner is ... text. Huh.

There are a gazillion ways we can send messages to each other these days: email, chat, your favorite social medium, send a postcard, make a phone call, walk over and say hi, etc., etc..  Some of these were the stuff of science fiction when I was a kid.  In particular, I think it's finally time to say that videophones are commonplace.  Most smartphones can handle it, and the bandwidth is there in many places, though certainly not everywhere.  Even so, millions of people have the ability to make a video call should they so choose.  Probably more like hundreds of millions.  And many do.

And yet ... if you have a video-capable smartphone and you want to send someone a quick message, or you're a celebrity and you want to let your fans know when your next appearance is, or you're a bank and you want to send your client a security code for logging in, or you're a wireless carrier and you want to send your customer a balance update, or even in some cases a spammer who wants to tell someone they may already have won a fabulous prize, or for any other number of reasons, what medium do you choose?  You send a text message.

This is really not all that new an idea.  In the 1800s, for example, people would send telegrams and cables, or -- in densely populated areas, at least -- dash off short notes for messengers to carry.  The diction is even strikingly similar to the modern equivalent, and it's even more striking given that there is massively more bandwidth available these days.  Clearly the problem is not that you have to crowd everything into a 160-character SMS message.  There are any number of ways around that.  Nor are you paying by the word, as in the old days.   With all the ways that one could send a message, right up to a high-bandwidth video connection, people are choosing to text.

What parameters might determine this?

Text has about the lowest bandwidth of anything that's in regular use for communication.  If you've ever heard anyone try ... to ... repeat ... what ... they ... were ... texting ... as ... they ... typed ... it ... in, you were probably gritting your teeth.  Even if you can text as fast as you can talk, with liberal use of abbreviations like OMG and U, it's still much more mental effort than just, y'know, talking.

As a side-effect of the low bandwidth, text is notoriously bad for conveying inflection and other nuances.  Emoticons only help so much.  Was that smiley sarcastic?  Is that frowny because of what they're telling me, or because they're telling it to me?  I texted them five minutes ago and they haven't replied.  Are they busy or do they hate me?  And so forth.

Text is so-so for latency and reliability.  Messages get dropped form time to time, or hung up in the ether for minutes or hours with no indication of whether they've been delivered or not.  Even under ideal conditions, you have to wait for the other party to type in their entire message before you get to see any of it.

Where text wins, I think, is setup time, which is as minimal as can be.



There are two main types of protocol: Packet-switched and circuit switched.  In a packet-switched protocol, the sender constructs self-contained packets and sends them to the receiver.  Since each packet is self-contained, individual packets may get lost or misdirected, and there is no guarantee that just because one arrived, any other will as well.  The prototypical packet-switched system is the mail, and to this day internet protocol documents speak of "envelopes" and "addresses".

In a circuit-switched protocol, the two parties first establish a connection (as we tend to call it these days), and then communicate over it.  Once the connection is established, messages flow over it in either direction (though in some cases they must take turns), until the connection is closed, either deliberately by the participants or by some sort of external disruption.  In general, you have some indication that this has happened, and if you do have a connection established, it's quick and easy to say "did you hear that?" or whatever if there's any doubt.

The prototypical circuit-switched protocol is the telephone.  When you place a call, you are establishing a connection.  Originally, the operator would use a patch board to set up an actual electrical circuit.  Thence the name.

Connections take a while to set up.  When you call someone, you put in their number, their phone rings, they stop what they're doing and answer it, and generally say "hello" or something to make sure you know the connection is established.  And then you talk.  A video call works much the same way, and for the same reason.  It's establishing a connection.

There are currently two widely-established packet-switched media: email and text.  I say "media" here because I'm talking about how things look to the people using them, as opposed to network protocols like TCP, UDP, ICMP and so forth, and I'm leaving aside services like Snapchat, which go beyond text, because it's early days yet.

Of email and text, text is much lighter weight.  Email more or less requires a subject line, and if a simple email evolves into a conversation, each piece of the conversation general contains everything previous. It's possible to have a rapid-fire email conversation, but it's a bit awkward. It's also considerably more likely that the recipient of your email isn't going to look at it for an indefinite amount of time.  For better or worse, if you're carrying your phone, you're likely to know immediately if someone has texted you.

Put all that together, and text wins, easily, on setup time.  If you already have a window open for your recipient (a sort of mini-connection, but without the overhead of setting up both ends), you just type.  And that's it.  Even if you don't, it's generally easy to pick a recipient from your contacts.  And then you just type.  And that's it.

Because the setup is so easy, a text can easily turn into a conversation.  If the conversation gets involved, you can always text "call me" or whatever and get the benefits of a real, higher-bandwidth connection but, crucially, this is opt-in.  You only pay that price if it turns out to be worth it.


It's now been almost twenty years since Kurt Dahl predicted that in the year 2020 -- then still comfortably far in the future -- there would be no need for kids to learn to read (See the Field Notes take on it here).   Instead, "text" became a verb, one used most by the very kids who would have seemed not to need it.  As always, it's easy, and pointless, to criticize in hindsight, though it might have been a clue that the prediction itself was conveyed via text.  Certainly there are many reasons why text should still be around, and texting is probably not a particularly big one.  Nonetheless, it's interesting that a medium that would seem to have so little going for it would win out, and that this could be due not so much to the virtues of text itself, as to the economics of communication protocols.