1. <<
  2. The BabelFish Blog
  3. >>
BlogEd 0.7
BlogEd 0.7 BlogEd 0.7 June 18, 2005 4:20 PM
After the last 8 months of serious work I am very close to releasing a new version of BlogEd (v. 0.7). I am publishing this entry here using the CVS version of the to be released code. Here are some of the improvements I can think of right now:
  • BlogEd is now based on a Semantic Web database (Sesame)
  • It can publish files to a server implementing the MetaWeblog API
  • It comes with a really cool wysiwig html editor kafenio
  • You can place pictures into the content
  • A lot of user interface improvements
  • A lot of bullet proofing (though not quite enough I am sure)
 There really is a lot more to do. I spent a lot more time than may seem reasonable to anyone not used to the pace of research, clearly setting the rdf foundations of BlogEd. I really have to thank Sun Microsystems and especially James Gosling, for the patience they have shown in allowing me to work at my pace. From here on things should move a lot faster.

There are still some major things I want to add to the foundations. I really need to get hold of a database that can do inferencing efficiently, and that can work especially with Combined functional properties and Combined inverse functional properties (CIFP). This is what I have found to be wanting most often in the current version of Sesame, and which has forced me to jump through a lot of unnecessary hoops.

Apart from that I need to add Atom support (which is just in the process of getting the final stamps of approval by the IETF). And of course there is a lot of work to improve the User Interface. I am very interested in seeing what can be done with the Netbeans Framework. In any event this could be where most of the improvements for the next version will go.

Ok. So now lets see if this works...

MicroPayments June 15, 2005 5:14 PM
note: I was reading a blog entry by Hal Stern on Distribution vs Compensation, and started writing a reply to it, when I noticed that my reply was starting to get just a little bit long, and so I thought it would be more apropriate to post this as its own blog entry. This thought has been bouncing around my head for the last 5 years, so I thought it would be time to put it out in the wild and see how well it fares.

I am really interested, though I have not yet had time to follow the debate or the state of the art concerning micro payments. I keep thinking a case can be made that part of the .com collapse was due to their being a gap between the minimum payment for an add ($5 per 1000 page views) and the minimum payment that one can make with a credit card (something around $10 + the cost in time of entering all the data). Between these two extreemes lies a gaping hole that excludes many forms of transactions that we have no trouble making in real life. Many items of our every day life can cost under $10: think a packet of chewing gums, an apple, shoe laces, etc... All these transactions have been excluded from the net. (note: I believe that the transaction cost of paypal is much lower, so this may undermine some of the thinking here). So during the .com 'bubble' the main currency was the add click. When the value of that currency collapsed, so did all the companies that relied on it.

Those projects that survived were those able to bulk up many micro transactions, or place cost in another part of the value chain. So Red Hat was able to displace the cost towards support, Google by growing the value of the add click, Skype by creating revenue opportunities by connecting their free network to the old telephone network, Sun now sells machines + software solutions thereby grouping the cost of both together, and Linux and Open Solaris grow by finding indirect value to their improvement (each individual actor's time in fixing or improving something could be thought of as a microtransaction to themselves (they get back what they put it)).

I find some further support for this theory in the fact that whilst there was a .com bust there was no equal telecomunication bust (the problems there have a very different origin, stemming in Europe from the huge price governments extracted from the telephone companies for the airwaves). In the mobile handset world, there is no microtransaction problem. Your phone call can be charged by the second down to the lowest cent.

So keeping in mind that it is not because a transaction is not monetised that it has no value (the growth of many free software projects belies this), it is clear that there may be some and maybe many types of transactions that have not been monetiseable due only to the technical difficulties in doing so. Part of the problem with aggregation is that it helps mostly build monopolies. For the larger the aggregate offered to one, the more value the currency one is using has (€ vs $?). Ebay keeps growing because so many other people use it that the network effect of using it strongly re-inforces its value. But what is the cost? Well think about smaller players such as The Economist. Writing and collecting articles is expensive. Because of the gap between the price of advertising and the cost of a credit card transaction, online newspapers cannot charge per article. You have to buy a subscription. The subscription is a bulk sale. But what if you don't want to buy the whole year's economist? What if you just want to buy a few articles, from time to time? This is even clearer with the Oxford English Dictionary. There the cost is $29 per month! But I am never going to use $29 of the OED per month. They are therefore loosing out the huge market of casual users they could gain who would have absolutely no trouble paying 1 cent per word lookup. There is no way I am going to buy the OED then as a casual user. I have to buy it as part of another package: perhaps MSN, or AOL? There you go. The big winners will be the super agregators. Not so surprising that the .com bubble collapsed with the creation of the Time-Warner AOL super aggregator.

But the technology for micro payments should be here now. Where is it? As a free software writer it would be really helpful if I could sell BlogEd releases for $1 or less a release per user. 100 000 users could help me grow the team of paid developers to work on it. And it could furthermore remain Open Source. Because who would bother trying to save $1 for a tool, by going to the web site and compiling it themselves? Anyone who could do that would be in a position to earn a lot more per hour than the cost to them of doing so: just think of the time to find the web site, look up the latest branch, read the mailing list to find out if this is the currently developed version, etc... You would need to be able to do that in less than 5 minutes for the saving of $1 to make sense. And then you still would not have complete trust. $1 pays each user the trust that there will be support behind the product they use. The open source nature would give them trust that my egoism would not get in the way of their trust - ie. I could get bored, but if the product had enough users, someone would udoubtedly take possession of the stream of money. The product would finance its own existence.

My next laptop?
My next laptop? My next laptop? June 15, 2005 11:20 AM
What types of benefits does Openess bring? Well for one it allows a technology to grow in many more ways than any single organisation can afford to explore.

So here for example is Niagara, Sun's next generation chip. It has 8 cores and 4 threads per core == 32-way computer, as explained by Jonathan Schwartz's blog entry The difference between humans and white mice. It is optimised to be a low energy, high throughput server computer, exactly the type of cpu that could save Google millions in energy costs alone. Tim Bray has been following this more closely than me, and has some more detailed explanations on how this affects programming in his recent post On Threads.

But as Tim notes at the end of his post, an interesting point about Niagara is not just that it is very fast, but that it consumes very little energy, a very important requirement for portable computers, and explains why those laptops have always been a generation behind in speed to their more hungry desktop computers. So could this be the chip in my next 10x more powerful laptop?

Clearly Sun is not a company currently with enough resources to spend on the complex task of making laptops for what would certainly be a very small market. A laptop with Solaris would not come with the beautiful User Interface that I have gotten used to with OSX. It is not something I could give to any member of my family. But it would be extreemly handy to have as a development laptop. 10 times the power is something I would not say no to, even if it means loosing some useability, as it would allow me to get a peek at what is to come a few years down the road, and having the advantage of being ahead of the curve, can make a very big difference to what decisions one makes. So there is a market for this. But if Sun can't fill it, will anyone else? I imagine Tadpole might, if it is feasible.

So what would I be able to do with the extra spare cpu cycles? Well perhaps it would be a lot easier to have a crawler running in the background, or a database indexing all my files, or perhaps this would just make it a lot more fun to run something like Looking Glass as a desktop. Or perhaps there would be no problem running a really good Semantic Web Inferenceing engine. It may help running huge compilations in parallel. Or perhaps running a test application server in the background with some serious load.

So what would I be looking for?

  • a 16" or 17" screen (with better resolution that my 17" apple)
  • a 160 GB hard drive
  • 256MB or 500MB graphics card
  • super fast IO
  • excellent connectivity (including wireless)
  • please a good mouse (the apple one really works!)
  • volume and screen lighting buttons and a functioning hibernate that really works. (I could never get these silly things to work with Linux boxes consistently)

Now as I understand Niagara is best at integer calculations, so would that not be a problem for a laptop with a high graphics requirements? Not I think (but I know very little) if one thinks of the graphics card as a separate cpu. Indeed this is what the Scout programming language apparently does. And it is also what OSX does, by moving all the graphics transformations into the graphics card with Core Image. Java could I imagine be adapted to do the same. So the speed limitations of floating point calculations on Niagara may not be such a problem after all.

So there is an example of how openess can allow a company to grow into markets by proxy, that it would never have been able to touch alone. The more open something is, the more trust these other players can have that their markets won't be taken from them, the way Steve Jobs once killed the Apple clones.

The Power of Sharing
The Power of Sharing June 14, 2005 9:46 AM
Buisness week has just published a very good overview on the rise of collaborative structures that are using the internet to shake up industries all over the world.

Many of their examples are well known, such as eBay, Linux, Kazaa, Gnutella, Wikipedia, etc... I had not realised that Skype was based on the same principle though, and I had not yet paid attention to many of the other examples they provide. It is mooving fast.

On that subject, I should mention that Sun has just released today most of the code for Open Solaris, their super resilient Operating System, under one of the freest licences available. This means that it should be really easy for anyone wanting to start a small company to tune that OS to any specific market. A Laptop running Niagara perhaps?

Apple OSX Tiger
Apple OSX Tiger Apple OSX Tiger April 30, 2005 1:29 PM
Yesterday, a couple of hours before the official release time, I received my pre-ordered OSX Tiger, the new release of Apple's Operating System. I installed it immediately - after of course making a full bootable backup of my hard drive with Carbon Copy Cloner onto an 80MB partition on my external 500GB hard drive. That bootable copy has now come in really useful: the Cisco VPN software that I need to connect to the Sun intranet does not work yet with the new release, so I was able to boot the old Apple Cube onto the external partition, and check my mail there. Always make a backup!

Well this is a really great OS update, worth every euro cent spent on it. The most useful feature is the new Spotlight technology which is finally going to allow me to make use of the 500GB of email I have accumulated over the years. I subscribe to a large number of open source project mailing lists because their public archives often have bad or non existent search engines. The YahooGroups search engine for example is so awful that I just shudder at the thought of having to look something up there.

Just as I finished the installation I heard that the latest version of java (java 5) was available as a download from their site. Installing it does not delete any previous versions, so I now have three great Java releases to test my software with. And they seem to run a lot faster. The latest Early Access Preview version of IntelliJ is much snappier using java5. It's as if I bought a new computer.

The above reasons were my main motives for switching. But there are many more changes that I was looking forward to. John Siracusa has written a very good write up of a lot of the new features in Tiger in a detailed 21 page article. He points out the good and the bad of Tiger with very detailed and convincing arguments.

I am uncle :-)
I am uncle :-) I am uncle :-) April 6, 2005 9:18 AM
My sister Christina early in this morning (6 April), gave birth to my nephew and to her husband Marcel's son :) Louis at the Triemli hospital in Zürich. Christina seems to be very well from the short conversation I had with her this morning.

From the photo I must say that Louis, a nationless baby does look amazingly cute. He weighed 3770g and is 51cm tall.

urn:isbn March 31, 2005 9:30 AM
My previous post contains a urn:isbn URI to identify the "Service-Oriented Computing" book. The isbn Universal Resource Name is the standard Universal Resource Identifier for identifying a book, just like "http" is the standard Universal Resource Locator for identifiying a web page.

But it does not work with either Mozilla or Safari. And so instead web page designers use the semantically less precise http URI to identify a web page that speaks about the book. This page may not be the one that interests the reader of the page, and it might not be the best page to get information about the book in the future. It also means that search engines have to do a lot more work to figure out what the link is pointing to.

So what is needed? Firefox should lead the way here. It should give a few simple mappings from isbn URNs to web pages, and give the reader the choice which one he wants to consult when clicking on that link. The user would of course be able to add or alter these mappings. When clicking on the Service Oriented Computing link in the post below I would then for example have the choice of either selecting the Library of Congress menu item or selecting the amazon.com menu item.


I have posted about this issue at the following places:
  • FireFox MozillaZine forum. This forum is probably too specialised.
  • as issue 288416 on the mozilla bugzilla page. Ah it was a duplicate of bug 269917 which needs a bit of support to get it on its way.
  • MozillaZine Firefox features group, which should be a better fit.
  • new topic on the Apple Safari discussions site.
    Update: Apple pulled it! Perhaps because I mentioned Mozilla? Or perhaps the moderator there just did not know what URIs or URNs were and thought I was attempting to sell books. God knows! There is no way of finding out, and no open way of complaining. I refuse to send bug reports to Apple as they don't have a public database to store them in, and don't provide any feedback on where the bug is in their process. Why should I be working for them for free, if I can't even get full acknowlegement for the work I have done. Too bad for them.


  • There were some interesting feedbacks on the sem-web irc chatlog
  • And Mark Baker left one of these dangerous thought traps: "[...]just like "http" is the standard Universal Resource Locator for identifiying a web page". No, http is like (void *), untyped. It's for resources, not Web pages. ISBN should have gone with http://isbn.org/num/82734983274. Oh dear. I can see now where this debate is located.
  • Tim Bray made a similar comment over iChat
Oh dear that is one of those endless debates,... But this example may shed some practcal light on it.

Of course Mark Baker is quite right in correcting me about the http uris pointing to web pages. I should have said that they identify resources that are usually dereferenced and for which we are usually returned web presentable presentations. If "http://isbn.org/num/827343274" identified a book record reliably, then it would be attractive to use that URI as the Identifier for the book. What I want is a stable, non changeable, persistent way of referring to books, without favoring any particular commercial institution, which everyone will agree on, and which will allow people to be clear that I am referring to the book. Currently the urn:isbn does the job.

A short fictive history
If I were to invent a short history of the debate I would say that early on in the web someone invented urns to point to things. Soon the IETF became deluged by people who had special things they wanted to name, starting with books (ok), to (journals) ok, to cars (oh no) to butterflies (oh god!) at which point the IETF probably shouted "STOP" and told everyone to use http uris as names.


A person named Takeshi made me aware of a little Firefox extension written by Piro that deals with ietf, issn, isbn and publicid urns. Really cool. It does not give you the option yet of choosing between different mappings.

Service-Oriented Computing
Service-Oriented Computing Service-Oriented Computing March 31, 2005 9:18 AM
A few weeks ago I followed a number of links on the topic Why Rest is Better and came across the Web Services Architecture document where I noticed that Francis McCabe was one of its editors. Frank has been working in the software Agent field for over 15 years and I have been in contact with him on and off since my year at Imperial College. In 1996 he had allready implemented the April programming language, a tuple based language that makes it easy to write agent oriented applications.

So I asked him if he knew if there was any good literature on the REST vs SOAP debate that would help me find my way through this complex space. He suggest I check out Service-Oriented Computing - Semantics, Processes, Agents which I promptly ordered and received a little over a week ago. This book is very wide ranging, covering in great detail it seems the whole landscape of web services, including agents, the Semantics Web, SOAP, REST, Peer to Peer, and much more. It is exactly what I needed. A careful texbook like (as opposed to marketing like) overview of this huge field.

I will report back when I have finished this.