13 Mar 2011 apenwarr   » (Master)

The strange story of etherpad

I don't actually know this story - certainly no more of it than anyone who has read a few web pages. But luckily, I'm pretty good at making things up. So I want to tell you the story of etherpad, the real-time collaborative web-based text editor.

I find this story particularly fascinating because I have lately had an obsession with "simplifying assumptions" - individual concepts, obvious in retrospect, which can eliminate whole classes of problems.

You can stumble on a simplifying assumption by accident, by reading lots of research papers, or by sitting alone in a room and thinking for months and months. What you can't do is invent them by brute force - they're the opposite of brute force. By definition, a simplifying assumption is a work of genius, whether your own or the person you stole it from.

Unix pipes are the quintessential example of a simplying concept in computer science. The git repository format (mostly stolen from monotone, as the story goes) is trivial to implement, but astonishingly powerful for all sorts of uses. My bupsplit algorithm makes it easy and efficient to store and diff huge files in any hash-based VCS. And Daniel J. Bernstein spent untold effort inventing djb redo, which he never released... but he shared his simplifying ideas, so it was easy to write a redo implementation.

What does all this have to do with etherpad? Simple. Etherpad contains a few startling simplifying assumptions, with a surprising result:

Etherpad is the first (and still only) real-time collaborative editor that has ever made me more productive.

And yet, its authors never saw it as more than just a toy.

My first encounter with etherpad was when Paul Graham posted a real-time etherpad display of him writing an essay. I thought it looked cool, but pointless. Ignore.

Sometime later, I read about Google Wave. I thought that level of collaboration and noise in your inbox sounded like nothing anybody could possibly want; something like crack for lolcats. Ignore.

And then, a little while later, I heard about etherpad again: that it had been bought by Google, and immediately shut down, with the etherpad team being merged into the (technically superior, if you believe the comments at that link) Google Wave team.

Moreover, though, a lot of the commenters were aghast that etherpad had been shut down so suddenly - people were using it for real work, they said. Real work? Etherpad? Did I miss something? Isn't it just a Paul Graham rerun machine?

I couldn't find out, because it was gone.

The outcry was such that it came back again, though, a couple of days later, with a promise that it would soon be open sourced.

So I tried it out. And sure enough, it is good. A nice, simple, multi-user way to edit documents. The open source version is now running for free on multiple sites, including ietherpad and openetherpad, so you can keep using it. Here's what's uniquely great about etherpad:

  • Start a document with one click without logging in or creating an account.
  • Share any document with anyone by cut-and-pasting a trivial URL.
  • Colour coding, with pretty, contrasting colours, easily indicates who wrote what without having to sign your name every time (ie. better than wiki discussions).
  • Full document history shows what changed when.
  • Each person's typing shows up immediately on everyone's screen.
  • Documents persist forever and are accessible from anywhere.
  • No central document list or filenames, so there's nothing to maintain, organize, or prune.
  • Easy to import/export various standard file formats.
  • Simple, mostly bug-free WYSIWYG rich web text editor with good keybindings. (I never thought I would voluntarily use a WYSIWYG text editor. I was wrong.)
  • Just freeform text (no plugins), so it's flexible, like a whiteboard.
  • A handy, persistent chat box next to every document keeping a log of your rationale - forever.
  • A dense screen layout that fits everything I need without cluttering up with stuff I don't. (I'm talking to you, Google Docs.)
  • Uniquely great for short-lived documents that really need collaboration: meeting minutes, itineraries, proposals, quotes, designs, to-do lists.
Where's etherpad development now? Well, it seems to have stopped. All the open source ones I've seen seem to be identical to the last etherpad that existed before Google bought them. The authors disappeared into the Google Vortex and never came out.

A few months later, Google cancelled the Wave project that had absorbed etherpad. It was a failed experiment, massively overcomplicated for what it could do. Nobody liked it. It didn't solve anyone's problem.

And that could be just another sad story of capitalism: big company acquires little company, sucks life out of it, saps creativity, spits out the chewed-up remains.

But, you see, I don't believe that's what happened. I think what happened is much more strange. I think the people who made etherpad really believed Google Wave was better, and they still do. That's what fascinates me.

See, upon further investigation, I learned that etherpad was never meant to be a real product - it was an example product. The real product was AppJet, some kind of application hosting engine for developers. As an AppJet developer, you could use their tools to make collaborative web applications really easily, with plugins and massive flexibility and workflows and whatnot. (Sound familiar? That's what Google Wave was for, too.) And etherpad was just an example of an app you could build with AppJet. Not just any example: it was the simplest toy example they could come up with that would still work.

I get the impression that the AppJet guys were pretty annoyed at the success of etherpad and the relative failure of AppJet itself. Etherpad is so trivial! Where's the magic? Oh God, WHAT ABOUT EMBEDDED VIDEO? WILL SOMEONE PLEASE THINK ABOUT EMBEDDED VIDEO? Etherpad couldn't do embedded video; still can't. AppJet can. Google Wave can. Etherpad, as the face of their company, was embarrassing. It made their technology look weak. Google Wave was a massive testosterone-powered feature checklist from hell, and Etherpad was... a text editor.

No wonder they sold out as fast as they could.

No wonder they shut down their web site the moment they signed the deal.

They felt inferior. They wanted to get the heck out of this loser business as soon as humanly possible.

And that, my friends, is the story of etherpad.


But I'm expecting a sequel. Despite the Wave project's cancellation, the etherpad/appjet guys have still not emerged from the Google Vortex. Rumour has it that their stuff was integrated into Google Docs or something. (Google Docs does indeed now have realtime collaboration - but it's too much AppJet, too little Etherpad, if you know what I mean.)

When I had the chance to visit Google a few weeks ago, I asked around to see if anybody knew what had happened to the etherpad developers; nobody knew. Google's a big place, I guess.

I would love to talk to them someday.

Etherpad legitimized real-time web document collaboration. It created an entirely new market that Google Docs has been desperately trying, and mostly failing, to create. Google Docs is trying to be Microsoft Office for the web, and the problem is, people don't want Microsoft Office for the web, because Microsoft Office works just fine and Google Docs leaves out zillions of well-loved features. In contrast, etherpad targeted, and ironically is still targeting and progressively winning despite the project's cancellation, an actually new and very important niche activity.

The brilliance of etherpad has nothing to do with plugin architectures or database formats or extensibility; all that stuff is plain brute force. Etherpad's beauty is its simplifying assumption, that just collaboratively editing a trivial, throwaway text file is something a lot of people need to do every single day. If you make that completely frictionless, people will love you. And they did.

Somehow, the etherpad guys never recognized their own genius. They hated it. They wanted it dead, but it refuses to stay dead.

What happens next?


Pre-emptive commentary

I expect that as soon as anyone reads this article, I'll get a bunch of comments telling me that Google Wave is the answer, or Google Docs can now do everything Etherpad can do, or please try my MS Office clone of the week, etc. So let me be very specific.

First of all, look at the list of etherpad features I included above. I love all those features. If you want me to switch to a competing product for the times I currently use etherpad, I want all that stuff. I don't actually want anything else, so "we don't do X but we do Y instead, which is better!" is probably not convincing. (Okay, maybe I want inline file attachments and a few bugfixes. And wiki-like hyperlinks between etherpad documents, ooh!)

Specific things I hate about Google Wave (as compared to etherpad):

  • It's slower.
  • The plugins/templates make things harder, not easier.
  • Conversations are regimented instead of free-form; you end up with ThreadMess that takes up much more screen space than in etherpad, and you can't easily trim/edit other people's comments.
  • It has an "inbox" that forces me to keep track of all my documents, which defeats throwaway uses where etherpad excels.
  • Sharing with other users is a pain because they have to sign up and I have to authorize them, etc.
  • The Google Wave screen has more clutter and less content than the etherpad screen.
  • Google Wave has a zillion settings; etherpad has no learning curve at all.
  • Google Wave wants to replace my email, but that's simply silly, because I don't collaborate on my email.
  • Google Wave wants me to live inside it: it's presumptuous. Etherpad is a tool I grab when I want, and put down when I'm done.
Specific things I hate about Google Docs (as compared to etherpad):

  • It's slower.
  • The screen layout is very very crud-filled (menu bars, etc).
  • It creates obnoxious popovers all the time, like when someone connects/disconnects.
  • Its indication of who changed what is much clumsier.
  • Its limited IM feature treats conversation as transient and interruptive, not a valuable companion to the document.
  • The UI for sharing a document (especially with users outside @gmail.com) is too complicated for mere mortals, such as me, to make work. I'm told it can be done, but it's as good as missing.
  • I can't create throwaway documents because they clutter my personal "list of documents" page that I don't want to maintain.
  • I have to save explicitly. Except sometimes it saves automatically. Basically I have no idea what it's doing. Etherpad saves every keystroke and has a timeline slider; anybody can understand it.
  • It encourages "too much" WYSIWYG: like MS Word, it's trying to be a typewriter with paper page layouts and templates and logos and fonts and whatnot, and encourages people to waste their time on formatting. Etherpad has WYSIWYG formatting for bold/italic/etc, but it's lightweight and basic and designed for the screen, not paper, so it's not distracting.
There are probably additional things I would hate about Wave and Docs, but I avoid them both already because of the above reasons, so I don't know what those other reasons are. Conversely, I use etherpad frequently and love it. Try it; I think you will too.

Update 2011/03/13: In case you would like to know the true story instead of my made up one (yeah, right; that would be like reading the book instead of watching the TV movie), you can read a response by one of the etherpad creators. Spoiler: they have, at least physically, emerged from the Google Vortex.

Update 2011/03/13: Someone also linked to PiratePad, which is a modification of etherpad that includes #tags and [[hyperlinks]]. That means they accomplished one of my dreams: making it into a wiki!

Syndicated 2011-03-13 12:05:58 (Updated 2011-03-13 23:05:02) from apenwarr - Business is Programming

Latest blog entries     Older blog entries

New Advogato Features

New HTML Parser: The long-awaited libxml2 based HTML parser code is live. It needs further work but already handles most markup better than the original parser.

Keep up with the latest Advogato features by reading the Advogato status blog.

If you're a C programmer with some spare time, take a look at the mod_virgule project page and help us with one of the tasks on the ToDo list!