Opera Unite: Web Untied

The most revolutionary new feature in Opera 10 is also among the oldest. In this case, while the teaser promising a reinvented Web may have been over the top, the hype is factual. Opera Unite is a revolution, among other wheels in motion.

My favourite headline has been El Reg’s Opera to take web back to the old days.

For decades Web intellectuals have railed against the client-server model, argued that it is too stale and authoritarian, had the server point of failure and couldn’t scale with the exponentially growing Web. Power to the distributed systems. Then Google came along and showed you can build a bigger server. The bigger the problem, the bigger the server park. Problem, any problem, solved. But way back in the CERN pioneering days the client was the server, the consumer was the producer. This grass root idealism didn’t survive the Web’s brush with success in the mid-90s, and when the revolutionaries no longer had scale on their side the revolution faltered, ending up with SETI searchers and UFO fanatics.

Unite is old within Opera as well, I couldn’t put a date on it but as a concept it would be Opera 7 to Opera 8 territory. Having a server on a phone generated some buzz on Friday nights, or maybe it was the beer. I have some very simple ideas for what I want from computing, among them I want to liberate data from the machines they are running on. Lars’ concept, which Opera Unite is based upon, takes the opposite approach, it gives the machines a human face. To me this is a little weird, Opera is after all in W3C parlance a user agent, not a machine agent, and a little wonderful as well. It hasn’t fully dawned on people yet, but Opera might unite not just them, but their machinery as well.

Concept is one thing, doing the work is something else. There were some very tricky problems to be solved for a browser-server solution to work in practice, and it was integrated with the widget framework, which has a number of advantages, ease of authoring and a common security model among them.

This leads us to now. What if we have a revolution and nobody comes? Even if beneficial for the Web, it is no guarantee that Opera Software will benefit from the revolution they have started, past revolutions and Opera’s past performance should give enough historical evidence of that. Worst case this might end up as Opera Hypercard, technologically interesting, even influential, but with little commercial benefits.

There are immediate advantages, obviously to run the servers Opera would have to be installed on the machines, and it can be an argument to switch browser. Longer term Opera’s benefit largely lies in letting us eat a bigger cake. As long as their slice of the cake is as small as today others will profit from the sweat of Opera’s labour, but browser vendors today are marked by coopetition, competitors with a strong incentive to cooperate, though some of the vendors have other agendas as well.

Seemingly going from the client to the server is a step backwards, after all some can recognize the Opera brand, many more the Firefox brand, but almost nobody can name the Apache brand, even though this is arguably the greatest success story of them all. But by turning a client into a client-server, and the server as well, you have peers. Opera Unite wouldn’t be an efficient way to distribute content, or even to represent me or mine, but it can be a flexible way to communicate.

Join the Conversation

  1. Great write-up with many valid points and questions raised. Thanks!But what is Opera Hypercard? I don’t remember such thing…

  2. There is little to add to the post, or even to FataL’s comment. I hereby second the question regarding Opera Hypercard.Wanted to let you know, though, your blog posts (well, the longer ones) are some of the most thoughtful ones I see here on MyOpera. Thanks for this post.

  3. I meant the “well-known HyperCard”. Like it Opera Unite has a “we don’t know what it is, but it is cool” aura to it. Since I was interested in hypertext before the Web hypercard is an interesting case for me. Architectural constraints ultimately doomed hypercard, part because it wasn’t distributed, partly precisely for being a hypertext system, partly for other reasons.It could be an interesting academic exercise to build a hypertext system on top of Opera Unite. One of the reason Tim Berners-Lee’s WWW took off was that he decided not to make a (two-way) hypertext system because he believed it wouldn’t scale. He was right, but we still suffer broken links because of it. A peer-to-peer system could manage to do so, even the popular nodes. There would be millions of links CNN’s servers would have to update, but they handle millions of request already.

  4. Anonymous writes:What is meant by “a hypertext system on top of Opera Unite”? (As opposed to, Opera Unite running the webserver app serving up HTML pages.)

  5. The Web can be seen as by far the largest hypertext system in the world, except that formally speaking it isn’t a hypertext/hypermedia system, and that is by design. The most obvious “missing feature” is that a hypertext link in HTML is not a link, but a reference or pointer. If page A points to page B, B is not aware of that, and if B later changes address to C, A will still point to B, and you will have a dangling pointer. Until manually updated to C the link in A will be broken. If B were aware of all pointers linking to it, it could send a Change of Address message to all those pages, allowing them to re-route their link from B to C. The user would never need to know of the change.Of course you could use Unite to serve regular HTML pages, but what is the fun in that? The power of Unite is that you could recreate e.g. HyperCard as it should have been the first time around by creating a WidgetCard widget doing all that pesky maintenance for you. As I said this would mostly be an academic exercise. Sir Tim was right, this two-way linking stuff isn’t really that important, and these days we Google our way to the new address anyway. There would be niches where this would be very useful, even essential, and if someone bothered to make a WidgetCard application (which would be a lot more than just two-way linking) I would use it.

  6. (A comment to wanted: spartacus, an opera unite web proxy for iran that somehow didn’t register there)Using Unite to disseminate information created elsewhere should be fairly simple, and it could even be possible to use a (non-United) widget like Twitter Opera widget and turn it into a Unite service. If it is just to spread the word, something extremely simple and safe could be the basis for the service.The Opera proxy would be no real problem, it is a convenience, but to make a service harder to block you should have a look at Setting up custom domains for Opera UniteThe honeypot will be a greater and real challenge, however there is literature that should help. To spread information in a hostile territory look up Byzantine fault tolerance for a start. Also, if you have a number of identical clones of a service, and the number of services telling the truth is higher than the ones telling malicious lies, the service itself could help discover the truthful ones and out the liars. An open protocol as Opera Unite offers will give no protection against detection, as TOR does, but can help spread information in the face of widespread blocking.

  7. Hey, you understand this.txthttp://my.opera.com/gabydewilde/blog/show.dml/4109234debate +/& more txthttp://my.opera.com/community/forums/topic.dml?id=283169After the exHTML thing we can now create formats that rely on adoption for survival.I felt really hurt where you wrote: :D”What if we have a revolution and nobody comes?”I initially thought it wasn’t even worth explaining.At least have a good laugh at my spex.


Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.