I broke news last night on LouisGray.com about Twitter enabling a new API feature, “in_reply_to_status_id”, to allow developers to tie replies to their original reply source. Immediately afterwards Evan Prodromou of Identi.ca added the same functionality to the Laconi.ca source code, making two of the most popular microblogging platforms, Twitter and Identi.ca, along with the already supporting FriendFeed, supportive of threaded comments. Immediately we saw Dave Winer implement a proof of concept example, and YooPlace also implemented it into their own code. Loic Le Meur of Seesmic, the owners of the Twhirl Twitter client responded as well saying he was reading the article and looking into it.
So a big question has become evident now that we can threaded replies. Is the ‘@’ really necessary in your replies to friends on your favorite microblogging service any more? A comment by Steve Gillmor on Identi.ca got me thinking about this earlier, and I think he has a strong point. In fact, I’ve briefly touched the subject before here.
The ‘@’ is mostly a Twitter-invented custom brought to the service by its users with some slight roots in bulletin boards and forums where threading was not possible. Users decided the Twitter service made a great communications tool and began replying to their friends’ posted statuses with the ‘@’ symbol. There was no other way because Twitter wasn’t expecting to be a communications tool. The popular Twitter clients like Twhirl and TweetDeck and even Twitter’s own web-based client started catching on, and separating those posts with usernames prepended by ‘@’ as “replies”. What’s odd is that the only thing they recognize as a reply is if the username is prepended by an ‘@’ – they take no thought as to the actual username itself, which really is the actual substance of who the user was replying to. In fact, FriendFeed users are starting to do this as well since it only has one level of threading and users can’t comment on other users’ comments. (Twitter and Identi.ca are actually one-upping Friendfeed with their recent announcements)
So while the “@”‘s were a custom, they really aren’t necessary to determine if a user is replying to another user. In fact, even today you can use an XMPP tracker like Twitter Spy and Laconica Spy and track your username and get notified when a person mentions your username, exactly the way “@” replies work. It was silly that the “@”‘s were required to be recognized as a reply in the first place.
Now, considering you can now actually track on the back end the entire hierarchy of a conversation via the API “@”‘s are even less necessary as before. I’d like to see the various microblogging clients start to ignore the “@”‘s and allow users to simply type usernames when replying to another individual, assuming threads aren’t in place. Then, once threads start to show in your favorite microblogging clients, even the usernames shouldn’t be necessary.
I’ve mentioned before that IRC works this way and most IRC clients will look at the existing list of users in a room and automatically detect the username and notify the targeted user if the message is directed at that user. Not only that, but the IRC clients actually keep a cached version of the users in a particular room and will even auto-complete usernames if you begin typing in the username and hit “tab”. That’s what I’d like to start seeing microblogging clients do so long as they’re going to be supporting a communications platform, and it should start with Twitter and Identi.ca themselves on their own web interfaces.
Then again, all this may now be moot with threading available. Oh, and don’t even get me started on hashtags. (Those should be handled by the API, not in the content of the message!) I think I’m going to try a new experiment of just not using the “@”‘s like Steve Gillmor does – anyone else want to join me?
Well, if I remember correctly, Twitter wasn't originally designed with replies in mind. It was because people were responding to each other with @username that Twitter took over the custom. A *bottom-up* innovation! It may not be technologically necessary, but society may not be so quick to abandon the practice.
Also @ is a symbol that is immediately associated with a response. If people say @vincent, I immediately know if they're talking to me, vs. @Jesse, where I know they're not.
The microblogging clients are still broken though by not looking at the
username itself. The clients are only looking to see if there is an @, and
*then* they look at the username. In reality they should be looking only at
the username and categorizing it as a reply. The @ should have no
significance in the technology itself. And again, once threaded comments
are enabled the username isn't even necessary.
Yes, the @ is still really necessary, for three reasons:
1. it addresses people: you can address other people than just the writer of the dent you're replying to
2. you can address one or more people without necessarily replying to anything/anyone at all, just starting a conversation
3. it is necessary for disambiguation: any word could be a nickname but you're not necessarily addressing that person (or even aware that it might be an existing nickname) when using that word
Marjolein 3 is the only one that really has any merit. For 1 and 2 I can
still address someone by simply posting their username and no @ symbol. For
#3 you'll have a few ambiguities, but the majority of mentions of a nickname
by your circle of friends will be actual nicknames – it will also make
people think twice before choosing a nickname that is a common English
word. This has been around in IRC clients for ages and no one has
complained. And again, that can all be solved by technology – there really
is no reason for the '@' symbol if the microblogging client just handles it
right. The @ clutters up Twitter and takes one character out that you could
be using in the 140 char allotment you already have.
Jesse
Jesse,
IRC is a poor comparison, because you can address someone only by starting a message with their nick. At least I know of no IRC client that somehow automatically shows who a message is addressed to other than by the fact a msg starts with a nickname, followed by a comma or colon. Everything else is ambiguous.
In Twitter, I found the necessity to start a reply with a @nick a poor and needlessly limited implementation of the idea – even a tweet starting with 2 @nicks is still a reply only to the first.
(You're also incorrect addressing with @ was somehow “mostly a Twitter-invented custom”: it's long been a wide-spread custom in forum and blog-comment discussions, a custom Twitter users simply – and quite logically – continued to use in a new context.)
The advantage of identi.ca's implementation is that you can address (without any ambiguity) anyone and “anymany” by @-addressing them anywhere in your post. Sentences can be more natural that way.
And, of course, the more people sign up, the greater the chance of ambiguity. Whether some word is a “common English word” is no argument here, of course – not all conversation is in English, and people may sign up who do not even know English: how could they “think twice” about that? – how would you know the difference between a nick and a word in /any/ language?
I really don't care that it takes all of 5 characters when one is addressing 5 people – and I'm an advocate of limiting to 140chars at the same time. There's no “clutter” in using @ – just clear intent that isn't present without it: without the @ in front of the nick of all you're addressing one simply can no longer see who you are addressing (if anyone at all).
You won't see me dropping any of my @s!
[…] Jesse Stay explores the potential demise of the @ on Twitter in more detail. Personally, I’m completely against […]
Jesse, the @ isn't just a link mechanism. It is a key Twitter gesture. It says to a person, I am talking to you now, and it also tells your followers that you value someone that they should possibly follow. Twitter behavior is subtle and fragile. Beware of unintended consequences.
Adam that's fine, but the technology shouldn't be determining if we use the
@ or not. It should be the users that determine that. The technology
should be looking at the username and let the users decide how they want to
address someone.
Marjolein I use IRSSI and it doesn't require a ':' or ',' following the
username. In fact it will highlight a message and notify me if my username
is mentioned anywhere in someone's comment.
'@' was made popular by Twitter – I did admit it was around before Twitter
though.
The common-English words are the only reason I can see a need for “@”'s. I
still think there can be ways around that with well-written technology
though.
Well the idea of a metalanguage to address explicit stuff to the machine is a good idea, the day we are able to create machines that parses natural languages we will get rid of a lot of stuff we use in metalanguages until then stuff like the @,# will be needed.
The @ helps to create a client as simple as it can be, the threaded discussions makes it harder to write a simple client that doesn't require much for the user to decide which thread they are replying. Think of the IM clients who doesn't have twitter/identi.ca specific guis and the whole gui is implemented on the bot who talks with you, so this mean that for threads the bot will require a lot more of instruction from your side. Our bots aren't that smart yet.
I've got a few concerns about dropping the @ sign. I'd like to learn more.
1 – Lots of nicknames are also nouns. How can you differentiate between the two without some indicator that the word is a nickname? (i.e. boat marina and @marina)
2 – How do you thread conversations over IM? Or SMS?
3 – Again, when replying to multiple people at once, threading doesn't apply, and I don't understand how it would be sure if I was intending “John” or “@john.”
4 – I'm a beginning programmer, so I have no idea if this is valid, but it seems like parsing entire messages for potential usernames across federated systems would be complicated.
Marina, I'm going to try to do a follow-up post this weekend to share some
thoughts I've gathered around your points. Simply put, I think the root of
the matter is all centered around the “what are you doing” theme. Anything
outside of that realm should be meta information of some sort outside of the
message itself – i.e. replies should be linked via meta in_reply_to_status
API calls. Users intended to see a particular message should be tagged, but
tagging should be done via the API and the client should provide access into
that API. Regarding IM or SMS, IMO those should be information gathering
tools, not information posting tools, as they are too hard to guide users
into using the system in the correct manner. I'll ellaborate in my post
this weekend with some concrete examples – the conversation is too
fragmented at the moment and I think part of that is the fault of the
microblogging platforms not providing the proper tie-ins to de-fragment the
system.
Jesse
I look forward to your next post. My one main comment here is that a tool that doesn't fully integrate IM and SMS is not the right direction … a lot of my most important messaging is done while on the road and using SMS.
I'm with Marina on this, also you speak like the fragmentation and the meta information is more a pain in the ass for users than it helps, simple tags like @ became popular just because it doesn't increase the burden/complexity to users.
I can't believe it took them so long to add this feature. Being able to browse through a thread is the #1 thing I didn't like about twitter. The conversations were just too fragmented.
I'm with Marina on this, also you speak like the fragmentation and the meta information is more a pain in the ass for users than it helps, simple tags like @ became popular just because it doesn't increase the burden/complexity to users.
I can't believe it took them so long to add this feature. Being able to browse through a thread is the #1 thing I didn't like about twitter. The conversations were just too fragmented.
[…] week I spoke about how the ‘@’ on Twitter and other microblogging services may no longer be necessary due to the addition of the in_reply_to_status_id API call on services like Twitter and Identi.ca. […]
[…] week I spoke about how the ‘@’ on Twitter and other microblogging services may no longer be necessary due to the addition of the in_reply_to_status_id API call on services like Twitter and Identi.ca. […]
[…] week I spoke about how the ‘@’ on Twitter and other microblogging services may no longer be necessary due to the addition of the in_reply_to_status_id API call on services like Twitter and Identi.ca. […]
[…] week I spoke about how the ‘@’ on Twitter and other microblogging services may no longer be necessary due to the addition of the in_reply_to_status_id API call on services like Twitter and Identi.ca. […]
[…] week I spoke about how the ‘@’ on Twitter and other microblogging services may no longer be necessary due to the addition of the in_reply_to_status_id API call on services like Twitter and Identi.ca. […]
[…] week I spoke about how the ‘@’ on Twitter and other microblogging services may no longer be necessary due to the addition of the in_reply_to_status_id API call on services like Twitter and Identi.ca. […]
[…] week I spoke about how the ‘@’ on Twitter and other microblogging services may no longer be necessary due to the addition of the in_reply_to_status_id API call on services like Twitter and Identi.ca. […]
[…] week I spoke about how the ‘@’ on Twitter and other microblogging services may no longer be necessary due to the addition of the in_reply_to_status_id API call on services like Twitter and Identi.ca. […]
I'm with Marina on this, also you speak like the fragmentation and the meta information is more a pain in the ass for users than it helps, simple tags like @ became popular just because it doesn't increase the burden/complexity to users.
I can't believe it took them so long to add this feature. Being able to browse through a thread is the #1 thing I didn't like about twitter. The conversations were just too fragmented.
Well the idea of a metalanguage to address explicit stuff to the machine is a good idea, the day we are able to create machines that parses natural languages we will get rid of a lot of stuff we use in metalanguages until then stuff like the @,# will be needed.
The @ helps to create a client as simple as it can be, the threaded discussions makes it harder to write a simple client that doesn't require much for the user to decide which thread they are replying. Think of the IM clients who doesn't have twitter/identi.ca specific guis and the whole gui is implemented on the bot who talks with you, so this mean that for threads the bot will require a lot more of instruction from your side. Our bots aren't that smart yet.
Marjolein I use IRSSI and it doesn't require a ':' or ',' following the
username. In fact it will highlight a message and notify me if my username
is mentioned anywhere in someone's comment.
'@' was made popular by Twitter – I did admit it was around before Twitter
though.
The common-English words are the only reason I can see a need for “@”'s. I
still think there can be ways around that with well-written technology
though.
Jesse, the @ isn't just a link mechanism. It is a key Twitter gesture. It says to a person, I am talking to you now, and it also tells your followers that you value someone that they should possibly follow. Twitter behavior is subtle and fragile. Beware of unintended consequences.
When using Identi.ca in Pidgin, which is not threaded by essence, it becomes very uncomfortable. Identi.ca within Pidgin will die…