Log in

No account? Create an account
Previous Entry Share Next Entry
Twitter, and why it needs an RFC...
Twitter isn't a website. This is Web 2.0 people, there's no such thing as websites. What you've got are people connected to portals connected to services. In fact, forget people, what you've got are content generators.

I feel dirty after saying all that, but it's not entirely untrue in this case. Twitter isn't a website, it's a protocol, a means of connecting users via a set of diverse applications all using a common database and interface. It's like NNTP, if there was only one news server on the grid - you have a protocol for pushing messages up, pulling messages down, and a few control commands to do neat stuff.

On blogs, there are those who complain that the simplicity of Twitter is what has made it so popular, and that any additional complexity will undermine its unique selling point. Well, as far as I can tell, your "tweets" are forever going to be limited to 1120 bits, and it's going to be up to clients to add any value on top of that. #Hashtags, @Usertags, these concepts are simply client-side interface add-ons. Granted, there are specific additional database concerns (for example the way replies are handled separately), but as soon as the search protocol picks up speed, even those concepts can be handled entirely client-side without any server support.

The simplicity is what enables complexity. Here, have 1120 bits, do whatever you like with them, and have fun. I've built a simple client/server authentication protocol that allows a person to validate their twitter credentials as supplied to a website. Website says "Hey, tweet this code directly at me, then I'll know that's really your username", and suddenly you've effectively got double opt-in.

Then I've got a little job that runs off daily statistics and tweets them. Or the jobs that tweet about popular posts in a community. Hell, you can run it like a command line interface - once you've got authenticated users, you could accept remote commands to reset your server, query a database, or play a description-light text based adventure game.

If the Internet can actually hold its attention span for long enough, then there's no reason why Twitter can't turn out to be the killer app for SMS and the mobile Internet. Of course, the Internet is a fickle place, and Twitter isn't there yet. If it could be decentralised, spread across servers that talk to one another like Usenet was, then it would just become part of the Internet.. Sites like "Plurk" are missing the point entirely by trying to replicate the functionality rather than extend it.

But perhaps there isn't enough money in that as a solution. And it would be so typical if the Internet's need for money were to kill an opportunity for it to become a better version of what it's really meant to be.

  • 1
(Deleted comment)
Yeah, only I got all happy because I thought somebody actually replied for real, and then I found out that it wasn't a real comment about what I was saying, and got this face :o(

You can't win ;o)

i got as far as the NNTP part of this rant and started pondering porn binaries over twitter

It could be done.... Into how many 1120 bit parts would you have to split your average jpeg?

but if it can be done we should try doing it via facebook stati

That's all already there really.

An RFC (in the traditional sense - a request for comments) isn't really relevant as the service has been implemented. Plus that level of documentation both isn't requred and would be exceptionally tedious.

What twitter really needs is a contract. Which luckily enough, it already has, in the form of the Twitter API. All it really needs on top of that is decentralisation and aggregation to be that killer app (and sites like friendfeed(.com) are already starting to build the aggregation functionality in that area).

Overextending Twitter would be very dangerous as the protocol (and it does have one) really is only suited to public domain information. It's really just the evolution / re-emergence of IRC.

At the heart of it it's just a simplified RSS reader (that aggregates VERY simple streams) with a handful of XML-RPC methods for replying. If twitter started allowing you to add "non-twitter" friends, then decentralisation would be complete instantly due to the inherent simplicity of application.

An RFC isn't really relevant as the service has been implemented

Well, the RFC remark was kind of a throwaway, I needed a subject line.. :o)

But still, what you ideally want is a set of protocol standards independent of any organisation that's implementing them, which is where RFCs are good and all.. For a single implementation where the standards are self-imposed, for "RFC" substitute "API" and possibly "Instruction Manual".

But if you wanted to deploy it as a technology with entirely redundant multi-carrier transmission of "tweets" (a la Usenet) then you'd want to pick something and standardise it. I think it's kind of close to that anyway, in as much as the message content is standard, and the in/out API is getting close to a finished product.. You'd just need a protocol for the sharing of new tweets across the various tweetservers you'd set up..

Though I do wonder if perhaps the need for things like multiple news servers has diminished as locality stop being a real advantage to data transfer speeds, etc.. Has the distributive model had its day for all but the most fundamental services (DNS, routing, etc.)?

Oh, oops.. I didn't read past "exceptionally tedious" before I went off on my own exceptionally tedious rant. Ha. Yes, API. All good. Decentralisation too, I mentioned that.. I should really have read your whole comment first - damnit :o)


I recon when they realise they can't make any money out of twitter all this stuff will naturally happen anyway :)

I've not looked into the API but I presume it's pretty simple and self documenting so people will probably just run with it. I've noticed in a few of the clients (well, when I say a few, I mean twhirl) that there are options to connect to competing equivalent services.

I'd imagine they'll just stick to Http as the protocol, it covers all the bases. I wrote a spec about a year ago for a project to act as a proxy for all of your personal data, like a personal aggregation server that you expose to your friends, but in the meantime friendfeed launched with about 50% of the functionality done very well anyway.

The jist was this

name.co.uk/identity/microblog (read tweet etc)

/identity acts as an openid provider for web-wide single sign on, while the other endpoints offer simple feeds or webservices to pull or push your publicly accessible personal data.

Gives you total control over your content and allows global interop.

Your friends or watchers subscribe to your public endpoints (wow, that sounds sexy...) and your identity portal either proxies data from a delegate site of your choice (say you like to tweet using twitter, then microblog would just proxy using the twitter api) or serves up it's own implementation. It allows you to move between providers whilst keeping your one identity sacred.

Everything'll flow in that direction eventually I'd hope / imagine. Personal control of your web facing data and a nearly static endpoint.

I recon when they realise they can't make any money out of twitter all this stuff will naturally happen anyway :)

You have more faith in the system than I do ;o)

Look at Facebook - when they realised they couldn't make money off it, they decided to beat it with a stick until money fell out of its own accord - initiatives like Beacon, for example, where they wanted to get some cash out of the platform without really caring about the impact on the users or the type of service.

I wonder if, in order to fund outbound Twitter in the UK, they might ask you to sign up to receive marketing SMS from third parties in order to subsidise the cost of transmission. I mean, it's the sort of thing you can imagine, right? Receive 20 tweets per marketing SMS you receive, etc.. Then suddenly it stops being a means of compensating for cost, and becomes a revenue stream in its own right with demographic targeting and potentially even with contextual triggers based on what you're tweeting about personally, or what your friends are tweeting about...

One of the pains of working in marketing is knowing how easy it is for people in positions of power to come up with these things and to imagine that they're actually a good idea.. Hell, I'm even wondering if I'm a genius right now, but I'm not - that's just what the marketing does to you.. You think of a way to get money, and imagine that it might work. Twitter can go the same way ;o)

I've not looked into the API but I presume it's pretty simple

Honestly, it took less than ten minutes to build a posting client out of cannibalised bits of my LJ posting client. Since you built one of those about nine years ago, I would say that Twitter is well beneath you ;o)

The reading bit took a couple of hours, but only because I found it easier to write my own basic XML parser than to try to learn one that PHP already has built in.. I recognise that this is not good practise :o)

I wrote a spec about a year ago for a project to act as a proxy for all of your personal data, [...] but in the meantime friendfeed launched...

Still, you're only a year behind the curve, that's not bad.. I keep coming up with amazing ideas and finding out that they already got done half a decade ago.. That's what I get for deciding to get out of the Internet game and stick with LJ for years, writing off Facebook and Twitter as fads rather than evolutionary progressions of online networking trends. And I probably still haven't learned ;o)

Ultimately, you're talking about a personal aggregator and delivery mechanism, right? So effectively your service subscribes to all your social media output, and then renders that in some aggregated format to other people, who use their own aggregators to combine your content with that of their other friends etc...?

Oh, haha, then I read on to where you basically said that that's exactly what it does, albeit potentially in realtime rather than batched as I may have implied. I need to get less enthused by what people say, so that I can actually finish reading stuff before I start to chip in with my own opinion... Massive attention span issues right here ;o)

I think the main problem with that is that it requires an additional site to run.. Bit too geeky for the mass market, because where twitter has a purpose in itself, and so does Facebook, your personal aggregator would be harder to sell to people who probably already have established networks on those individual sites and don't see the point of necessarily combining them.. Not to mention the fact that a lot of those sites are interoperable in their own right - people posting their Tweets to LiveJournal automatically, or automated tweeting when they update their blog, or whatever.. It becomes increasingly difficult to sell something that enables true interoperability when a cheap hack provides 75% of the benefit already.

And thus the Internet tends towards spaghetti.... :o\

I always imagined your personal identity service to be one of two things.

The first being the nerd driven homebrew portal that lets you enter rss feeds as delegates and acts as an open id provider.

The second as a hosted service.

I think it'll get more interesting if people put up a good resistance to the current "cloud" (I feel dirty typing vapor-words, and bad puns) storage trend. I'm really hoping that people are going to want distributed computing style services and hosting ("access your files, anywhere") but a proportion of them will be conscientious enough not to trust a random 3rd party with their data.

Thus the home server is born. It's your identity, it's your remote access, it's you, online. It's hosted in your house, you can turn it off if you want to. Everybody (in internet terms) has a DSL router these days, give it 20 years? Probably some evolution of wireless broadband will be more common.

It's part pipe dream, part utility and part just a safer version of what we have now. It'll probably gain a little more impetus as people start to become wary of the Google-nanny-state (anyone that drives 95% of the webs traffic, and that's a conservative estimate, is in too much of a position of power to own any MORE of my data than they already have).

I think friendfeeds implementation is very very good. I'd like to see them offer openId delegation, but they're nailing the aggregation. I'd love it to be open source but I'm a realist (open source != good by default).

I don't think people will manage to monetize on a per-message basis anymore. People already realise that that stuff doesn't really cost anything to run anymore. As mobile phones and the like start to effectively go un-metered they're going to really struggle to justify per-message costs on the web, where everything has been free-to-the-consumer (at least on the surface) forever. The obvious model is your standard non-invasive contextual ad. They're prone to all the disadvantages for the advertiser as ever, but content has to be king in that regard.

It's a tough one to turn a profit from, but it's not like anyone ever made a profit out of IRC and except Giganews and pirates, usenet. They might have to run with the kudos that they've invented (almost) a new communication technique for the internet, in parallel to email and instant messaging (it's just "open" IM after all).

I suspect a lot of these trends to happen and have the people trying to make money pulled kicking and screaming forwards, ala digital distribution. Pretty much because the implementation is simple, someone will get round to doing it.

The web is actually getting better at interop, slowly but surely. More people are accepting things like OpenIds, syndication / aggregation and microformats as regular operation. I think there's still tonnes of cash to be made but people need to really try stay away from the old models to succeed.

I say this to everyone, but: dood, have you read Nick Carr's The Big Switch? I know how much you love validation/vindication ;) IIRC, he doesn't talk specifically about Twitter because he probably couldn't get solid time with it before the book was published, but really cool stuff in there about Web 2.0 and the absence of websites, the changing structure of the 'net and, mainly, the shift to data information becoming a utility.

Not as yet, but I suspect that I might need to in order to complete my goal of finding some way to get the Internet to give me some money in some way :o)

  • 1