Peter Cranstone, while pondering how the Google phone will deliver ads to its users, says that Google will have to do something similar to what Opera does with Opera Mini -- transcoding web pages -- for the Google phone. He adds that once Google gets around to doing this it will beat the crap out of Opera Mini, which probably won't find much agreement with Russell Beattie, who argues that someone should buy Opera just for the traffic that is now routed through Opera Mini.

What both gentlemen are probably not aware of is that Google already has a transcoder that converts pages into mobile-formatted on the fly. Now, rather strangely, the interface is not available anywhere as a start page as far as I know. Google does serve you a mobile-specific Google.com page depending on your User Agent, but the links that are delivered in the results page do not use the transcoder.

The only place where you can see it is if you use the mobile version of the Google Reader. In the entry-level screen on Google Reader, there is a link that says "see original," which can also be accessed by pressing '0' on your mobile phone. To access any normal page on your desktop browser via this transcoder, all you have to do is to append the URL you want to browse to the following URL: http://www.google.com/gwt/n?u=. For example this blog can be accessed this way: http://www.google.com/gwt/n?u=http://fatalerror.wordpress.com.

Currently, the transcoder supports most standard HTML, including forms, which means that you get to access things like email on the go even on a very low-fi handset, and also that Google gets another bit of your personal information (did I hear the privacy paranoid let out a collective gasp there?) for it to index and profile. The good part of the story is that it refuses to transcode secure URLs, which I remember was not the case with Opera Mini.

Now, here I also have to admit here that Opera Mini does a stellar job, but it also has a problem that you need to have J2ME support to be able to use it. Besides, the Google transcoder seems to be considerably faster while transcoding and rendering pages. For all you know, Google maybe licensing Opera's technology to do this (imagine: Opera Mini kills Opera Mini. What a headline!), but from what I remember Opera is running a mightily hacked up version of the Opera browser as middleware to make Opera Mini possible, while Google's approach seems to be in line with the more standard HTML Tidy/HTML Cleaner/HTML Parser/Tagsoup approach to de-mucking web pages, albeit a monstrously hacked version of it.

Updates have been far and few in between here due to the same old reasons: life being mostly all work and very little play. There are a couple of pretty interesting developments that has been in the works, I will write more about them if and when they work out. Meanwhile, in the technology sphere, other than the recent and continuing dalliance with Lucene, Nutch and crawling the tubes, there is one bit of technology - RDF and the semantic web - that's been taking more and more of my thought cycles.

Firstly, I will readily admit to not yet understanding the core concepts -- triples and the subject, object and predicate soup -- to the required level of finesse, but I am trying hard to implement a subset of it in regular and existing applications, so that data can describe itself and generate multiple views that would otherwise be pretty much impossible to. This is also necessary because most of the tools in the RDF space -- including application frameworks and data browsers -- are far from being scaleable or stable enough at this point. Besides, paradigm shifts are best left out when it is based on currently evolving technology.

Over the past week I've been putting the finishing touches on version one of an internal API and remodeled it to be REST-compliant than to use XML-RPC as the earlier one used to do. It took me a while to wrap my head around URIs being used as unique identifiers and other concepts, but you can call me a convert now, after having seen the positives, thought I will admit that getting the URIs right is one hell of a bitch.

The time has finally come for me to get my hands dirty with a piece of technology -- Java -- that I have stayed away from. The week has started bright and early with the lovely NullPointerExceptions, servlet errors and other bits of incomprehensible language that I am getting my head around. No wonder the Java web application deployments have the .war extension, it is almost a war getting them to work at times. That said, Nutch and Lucene is one impressive bit of technology, so is the fact that Zend Search is binary compatible with Lucene indexes. It is not wonderful to have the best of two different worlds on the same platter?