Wednesday, 17 June 2009

Why I think Opera Unite is missing the point

I watched the PR fuss over Opera Unite with some interest but I do think that they have missed the point. This posting by Chris Messina raises some valid points about the use of Opera as a proxy but I think that this covers only some of the weaknesses.

Over the years, I have seen proposals for servers running on individual personal computers and on mobile phones and they consistently look to me like a solution looking for a problem.

Technically, it is easy to create or configure an HTTP server on most devices. On a PC, a server like Apache is not hard (for a techie) to install and configure although I don't see average users trying it. It is not hard to port Apache to smaller platforms but I have also seen (or created) custom HTTP servers developed for more embedded platforms. Creating a basic HTTP server in C/C++ or Python (or other languages) is easy. All of the clever options that 'real' web servers support are more challenging but probably not required for smaller devices.

What is clearly more challenging is deployment and configuration for average users, but this ties into the user cases. Given the assumption that it is practical to create an HTTP server on a range of devices, the question arises of what to use it for and how is the address made accessible. I think that both of these are real problems. I think that Unite addresses the deployment and configuration challenges but still does not come up with good use cases.

Opera try to handle publishing addresses by means of their own proxies. The comments in the blog above show some of the problems with this but I don't see a good alternative. Unlike standard web servers that have relatively stable IP addresses and that are hooked into the DNS system to make URLs usable, personal computers and mobile devices normally do not have fixed IP addresses. They commonly use DHCP to get a transient IP address. This means that it is difficult for a remote client to know what address to connect to. Also, PCs and mobile devices are not necessarily on-line the whole time (but more of this below). In practice, a personal or mobile HTTP server would need to get an address and then push its address to some central addressing server so that clients could find the address and tell if the server is online.

When we come to consider what should be published, I think that Opera are just wrong in their assumptions. I think that it is daft to try to publish static content such as images or other media files from a PC or mobile device compared to using a hosting service. If a user wants to self-publish then they have to consider issues such as backups and access restrictions. I seriously doubt that most users will want to do this. Why not just push the content to Flickr, Facebook or one of the groups services? Even if you don't fully trust the hosting services' backups they will provide basic access control and the content will be available approximately 24/7. If you want to retain control over the content then set up a cheap hosted web site. If you want to exercise access control then consider email or set up a proper site! If these files are made available from your home PC then you will have to effectivekly upload them whenever they are accessed - most home accounts are not great at uploading so this looks like a bad idea to me.

If you do want to publish any static content then surely you want the content to be available most of the time. I don't know about you but my laptop and smartphone are not connected 24/7. They are offline or turned off for substantial parts of each day.

The Fridge and Lounge services in Unite look a bit me-too to me. I could knock up such a service on a hosted site very quickly and they would be available when my home PC was turned off. I haven't looked but I would be very surprised if this type of service was not cheaply available with various hosting services. I guess that Unite makes it available for free and with less work than setting up a hosting service - you trade increased convenience for the price of going through Opera's servers. As an alternative, try setting up a closed Yahoo group and you can get many of the benefits.

I do see value in a local HTTP server but for more specialised purposes:

  • If a device (maybe embedded or mobile) wants to make its status or other dynamic information available then HTTP is a good enough protocol to use as it is well understood. I could see the value in a service wherby such devices ran a server and then informed a central service that they were available for information retrieval. This is a pretty closed use case.
  • Specialised HTTP servers such as synchronisation servers are useful but they will tend to run when required and have specialised methods for making contact with a client (you can tell that I used to work on OMA Datasync).

So, although I like the idea of running a local HTTP server on a PC or mobile device (it is quite fun to consider the porting issues), I still think it is a solution looking for a problem. The Opera site mentions future developments and the ability for developers to work with it but, heck, if I want to create web services, I will play with a hosted service. I can use Django (just one of many) to build something and all that Units gives me is some convenience at the cost of beling locked into Unite. I would love to be wrong but I don't think I am.

Having mentioned Chris Messina's post above, I subsequently came across a response from Lawrence Eng here.

I should add that I am not criticising Opera for not going open source
and I am not commenting on the technical implementation - I have not
spent the time to look at that so I am prepared to assume that it is
fine. My view is that the whole concept is a problem looking fr a
solution. Some of the blog comments make the point that it is easy
for naive users to use and this may turn out to be sufficient.
Alternatively, somebody may come up with the killer use case that I
have overlooked. until then, I will rem ain interested but sceptical.

No comments: