Co-linux is a custom linux kernel that can run as a windows application. It is bundled together with a debian linux base distro. On a whim I tried it today and I have to say that I am impressed. It boots very fast. Once booted you have what is known as the debian base image. 2d graphics are not implemented on colinux. But since linux guis can be served over a network that is not a problem. So rather than emulating some crappy display driver you just do apt-get install vncserver, download a vnc client for windows and tada graphics.

The rest is just straight debian configuration. For the average windows user that is pretty hard of course. But been there done that so no problem for me. I’ve been at it for a bit over an hour now.

The hard part was convincing windows to do internet connection sharing and remembering how to configure networks in debian (it’s been a while so it took me a few google attempts). After that it’s apt-get this and that. Woody was obsolete the day it was released years ago so I fixed sources.list and did a apt-get dist-upgrade to upgrade to testing. Then a few apt-get install commands to get an xserver, kdebase and vncserver (this is all explained in the co-linux documentation). Then I started a vncserver and connected to it using tightvnc (a nice vnc client for windows) and I am now looking at a kde 3.2.2 desktop. It’s actually running at native speeds. The only bottleneck is vnc so graphics performance basically sucks. I’m going to try using the cygwin xserver as well.

Berlin photos

A few weeks ago I spent a weekend in Berlin with my father and nephew. We made lots of photos and I have put a small selection online.

qemu & knoppix

Recently knoppix 3.8 was released. Knoppix is a nice linux distribution that you can boot straight from cd. It has two primary uses: showing off some linux desktop software without actually installing linux and doing maintenance on pcs (by bypassing the installed os). Fairly nice but useless to people like me since I am not a system administrator and know how to setup linux. No, the reason I dowloaded it was qemu. Qemu is a computer emulator that can emulate a variety of processors and is apparently advanced enough that it can run stuff like windows, linux, macos X and a whole bunch of other operating systems. Now that is interesting. And some guy put a nice qemu/knoppix bundle up for download.

Of course the problem with emulating is that it is slow. And it shows. Booting into KDE from the iso took a whopping 30 minutes. Then loading in konqueror took another two minutes. But the point is that it seems to work. With a ten fold performance improvement it would become quite useful. Of course stuff like vmware is much more suitable for doing this kind of thing. But the nice thing about qemu is that it is not limited to just emulating x86 but can also emulate a mac, a sparc and probably any processor architects that the qemu developers choose to support. It’s a tool with lots of potential and I expect to hear a lot more from it when it matures over the next few years.

no more referrers

Apparently the pornsites have found my site and are now unleashing their stupid tools on it in order to get their links in my referrer thingy. I’ve been getting thousands of hits from domains with such words as ‘rape’ and ‘perversion’ in them. Presumably this is intended to improve their google rankings. Anyway I don’t feel like monitoring referrerstats anyway.

So bye bye referrer thingy.

NVU 0.90

NVU, pronounced n-view ( is an evolved version of the mozilla composer tool in the recently discontinued mozilla application suite. I’m already a thunderbird and firefox user. So I’ve followed the very rapid development of NVU over the past few months. Every month Daniel Glazman produces a new version and with version 0.90 last week it seems well on its way to become a powerful tool for anyone interested in writing html without touching code. While I like messing with HTML and CSS, I’m pragmatic enough to not do this whenever I am writing content. When I write content, I don’t want to be distracted by technology. Therefore, tools like NVU are a potentially interesting addition to my toolset.

I’ve already used the 0.70 and 0.80 versions at work for editing changelogs and other release documentation. I was not impressed that much with the versions until 0.80. There was a particularly annoying bug with newlines and a lot of rough edges. 0.90 fixes the bug and provides a new polished look and feel. Suddenly it feels like a real application. It’s amazing what a bit of finishing touch can do for an application. It still needs some more of that but feature wise it has become a very nice application:
– support for all important html tags
– css editor
– support for layers and tables
– support for templates (ala dreamweaver)
– integrated site manager
– extension support

Extension support could actually become a killer feature. Many firefox development extensions should be portable to NVU. If that happens, dreamweaver suddenly looks like a toy.

I just loaded my IE unfriendly, homepage in NVU and to my surprise it looks fine (dreamweaver can’t handle it). All of the css positioning works as intended, the fonts are as specified and you can just click and type to add text. Combine this with the template functionality and you have the perfect tool for sites with elaborate/complicated designs.

Netbeans 4.1 Beta (2)

A few days ago I commented on my intention to try out netbeans. Sadly due to a bug in the project creation wizard, I didn’t get very far and at this point fixing things manually (by hacking the project xml file) is not something I want to do. I did manage to get a freeform java project going. However when you want to edit jsps in such a project you need something called web modules. Probably a web module is nothing else than an extra set of tags in the project xml. However, the gui doesn’t really allow you to add one by other means than using the web project wizard. And this thing crashes when you try to finish the wizard. Considering that this is a beta, I don’t think this is very bad.

Performance on the source only project was really good, by the way. I added a large project with multiple source folders and everything remained very responsive. Eclipse really chokes on this kind of project. Netbeans seems to take a more lazy approach to validating: if you don’t open the file, netbeans won’t do anything with it. Eclipse on the other hand tries to pre compile as much as possible and will tell you if there’s a problem with a file you are not editing (remove a jar file from the classpath and red error icons all over the place). While this makes netbeans very fast it also makes it less useful than eclipse: I really want to know when breaks because I changed something in Of course you can manually validate the whole project simply by building it.

Additionally the code editor sucks compared to eclipse. If code editing is not your thing you won’t care but what the hell are you using an IDE for then? The fact is that the eclipse java code editor is a more powerful tool.. The eclipse code editor is probably the best feature in eclipse. Netbeans compensates by doing everything else really well and by being fast and more scalable.

There has been a lot of discussion on the forum on eclipse performance. Several myeclipse, eclipse and netbeans people bothered to respond so it is worth a read. Bottom line is that the eclipse people are now working on performance (instead of features) and the netbeans people are getting rid of the bugs in netbeans 4.1 to make it ready for release. Both IDEs should be ready in few months. Probably netbeans 4.1 will be released a few months before eclipse 3.1. However, what really matters is the eclipse web tools project which is due this summer. Only when that and eclipse 3.1 are both ready, a fair comparison can be made.

Due to the problems with the web project wizard, I can’t give a full review at this point. I will try the release candidates when they come in a few months. Preliminary conclusion is that netbeans is a very capable tool that scales very well with project size. A severe limitation is the code editor which is a rather weak offering when you compare it to eclipse.

Scientific content should be free

A recurring topic on slashdot ( and in the scientific community is open journals: peer reviewed scientific journals that make their articles available for free. Today, slashdot commented on the ideas of the IEEE to maybe open their vast electronic library to the public. I am a big proponent of this and sincerely hope that they will do this. However I am very critical of the discussions about the cost. These discussions appear to be influenced heavily by publishers who continuously try to make it appear that these costs have to be very high. They propose that authors cover the cost of 3000$ (!!!!) per article.

This is where I disagree because as far as I can see these costs don’t really exist (or rather have to exist). I used to be a Ph. D. student. I wrote articles, submitted them to conferences and journals. I also peer reviewed articles for journals and conferences. I never received a single dollar for this work from the publishers and worse now have to pay to get access to my own articles (well I cheated by saving a copy).

My point is: all the relevant work in the publishing process is done by volunteers like me. Worldwide, scientists write articles for free and review other scientist’s articles for free. The only scientists who receive money from publishers are editors who, in some cases, get a modest compensation for their precious time from a publisher who makes a lot of money. I’m convinced there would be plenty of people willing to donate their time to do this. I’m one of those people. This work mostly consists of taking decisions what to publish and what to reject, organizing and coordinating the review process, etc.

Historically we needed publishers to distribute the peer reviewed articles to libraries and this is why publishers have enjoyed an enormous revenue stream for centuries now. The profits made by publishers are huge (billions of dollars). They continue to be huge because scientists need to publish in their journals because of the journal rankings (which are based on references to articles).

Now that we have the internet, this is no longer true.

Well not entirely. Of course you still have some hosting costs, site maintainance and maybe a bunch of people coordinating the whole review distribution process, content management & site maintainance. My point is: the per article cost of the whole process is extremely low. It’s nowhere near the amount of dollars named in the article. I’d be surprised and shocked if it were more than a few dollars. An organization like the IEEE should be able to fund this using sponsoring, advertising & volunteer contributions.

Of course they’d have to reorganize how they work. A journal is a periodic bundle of articles intended for paper distribution. Electronic publishing is instant (not periodic) and essentially free of cost. Beyond organizing the process and hosting there is virtually no cost. The process which is currently optimized for paper distribution is therefore obsolete. You need volunteer authors, volunteer reviewers, volunteer editorial boards for specific scientific audiences, supporting staff and hosting (here are some real costs) and a means to establish article and editorial board rankings (this is mostly a technical problem).

Editorial boards consist of key members of a research community who invite other scientists to contribute articles and do peer reviews. The output of an editorial board consists of peer reviewed articles. Not for profit organizations like the IEEE can take care of the editing and hosting. This will require some funding. Funding is available from sponsoring, advertisements, research funds, universities, society memberships etc.

Considering the amounts that are saved by taking publishers out of the equation, this should be no problem. Universities would save millions if the whole scientific publishing community would adopt this model.

So, IEEE, ACM and other not for profit scientific organizations: do your members and the scientific community a favour (and isn’t this what you exist for in the first place?) by making content available freely. There’s no shortage of scientists willing to do the writing and reviewing for you (I’m one of those people). The rest of the process can and should be optimized for online hosting. The costs involved with the latter part should be very modest. The benefits are enormous.

Netbeans 4.1 Beta

I have some mixed feelings about Netbeans. Long time ago, netbeans was my favourite IDE. This was during the days that IDEs cost money, were slow and generally a pain in the ass. Up until netbeans my preferred environment was an editor, a dos box and a compiler. Netbeans fixed that. The 1.x versions were quite nice and feature rich compared to for example Visual Cafe or Borland JBuilder though clearly netbeans was the underdog and probably not very suited for larger projects either. Then SUN took over and an uninspiring batch of releases followed. The most disappointing was the 3.x series which basically confirmed that SUN was out of touch with developer reality (it didn’t fix the performance issues, it didn’t add refactoring, it kept the awkward files system mount thingy, etc.). So IBM came along and kicked their ass with eclipse 1.0, 2.0 and currently 3.0 (and soon 3.1). Basically people jumped on the eclipse bandwagon because it was the best product.

To be fair, netbeans never had much of a community. Before eclipse the IDE market was dominated by Borland, IDEA and some others. And then eclipse came along and now dominates this market to the point that most former competors are joining. Even mighty Borland has joined the eclipse project. In summary, eclipse now has lots of momentum and netbeans never had much momentum.

Yet, Netbeans is not a bad product. I (re)tried it tonight and the 4.1 beta is pretty nice. It’s fast, responsive and seems to do all the basic stuff pretty well. On top of that it is loaded with webdevelopment features that eclipse doesn’t have. And to top it off, it has very flexible project settings and can (so I’ve heard) even import eclipse projects.

Its weak spot however is Java code editing. Eclipse is extremely good at this and netbeans isn’t. Eclipse offers you autocompletion, refactoring, source transformations, templates, quick fixes, on the fly compilation and much more. Netbeans only does a small subset of this (about half the refactorings, hardly any quick fixes), impressive still but eclipse is far ahead in this respect.

However, eclipse has a major problem: it is SLOW. It has severe problems scaling to modestly sized projects. And that is why I decided to give netbeans another chance. I already know the 4.0 series is fast. On my previous work PC (a pentium III 550) it had no trouble managing the projects that currently seem to give eclipse lots of trouble on my newer and much faster desktop (a 3GHZ p4). I won’t make the switch probably because of the code editing issues and because at work the standard environment is eclipse (this is unlikely to change soon). However, I want to know what I’m missing.

Check back for an update in a couple of days.

MP3 Album Art Tool

I have a well organized mp3 collection and stumbled upon MP3 Album Art Tool (, a tool for automatically retrieving album cover art for each album you have. Together with this winamp plugin:, you get nice cover art while listening to your music.

Of course in practice it doesn’t find everything but I consider a 60% score on my collection quite excellent since it includes a lot of stuff that is pretty hard to find anyway. The tool saves a nice folder.jpg in each album folder which means that windows explorer can be tricked into using the album art as a folder icon :-).

Open Office 2.0 Beta

Open office 2.0 is out and here’s a little review on the features that matter to me. That excludes all components except writer.

Lets start with my main problem with this release: they didn’t fix cross references. I’ve complained on issuezilla about this since before 1.0. Over the years I have grown convinced that either they don’t understand the problem or are not interested in fixing it (which would be the logical result of misunderstanding the issue). Anyhow, crossreferences are still badly broken. This means I won’t spend much time with open office regardless of its other qualities since crossreferences are very important to me (far more important than say word compatibility or shiny features I won’t use). In a document I typically have numbered stuff such as literature references, section, figures, tables etc all over my documents. Open office doesn’t offer an easy way to do refer to such items and editing them manually is not an option for any kind of structured document (e.g. scientific articles, legal articles, phd & master thesises, etc.).

UI. The user interface seems to have improved. But then it was really bad in version 1.x. It still looks like it doesn’t belong on the OS you are using (regardless of the OS). But lets be honest, things have improved. And of course SUN has a way of making enduser software feel awkward like no other software company. Their marketing department has clearly been all over the ui and tried to make the engineers hide all the non nativeness. So they’ve duplicated many of the office 2003 features but they remain bad copies that are clearly different in both look and feel. They’ve ditched the one application for everything paradigm (good); added dragable toolbars (no contextmenu for customization though). The application icons are still very windows 3.1 like (the commercial version probably will fix that). This is where the improvements stop, the rest of the software is still a usability nightmare. Some examples: the cross reference dialog (just open it and see for yourself); the task pane (improved but still awkwardly crowded with incomprehensible icons); the options pane (yikes!). I’ve said it before and I’ll say it here: Sun does not understand end users and it shows in all their software products with open office as the best example.

Imports. Still not perfect but much better than version 1. I work in structured way using words features more or less as they are supposed to be used (I don’t use new lines for whitespace, I use crossreferences for numbering, I leave figure positioning to word, I use paragraph styles for formatting, etc). Given that, you’d expect more or less perfect results when importing documents in OOO. Well at least the results are usable (i.e. the layout is not mangled like in 1.x). The images appear where you’d expect them. But again crossreferences are a problem. In all my documents, the crossreferences do not import correctly and lose their formatting. For example my literature references are a numbered list of references at the end of my papers. Inside the papers there are countless refs to the list items like foo [22]. Open office loses the brackets and with no obvious way of making the references there’s no way to get them back. It’s broken and you can’t fix it. And then there’s the strange behaviour of page numbering: I had no page numbering before the import but OOO ‘fixed’ that?! Thanks but no thanks. But credit where credit is due: the word import is very fast and overall more correct than 1.x.

Installer. Sun still hasn’t figured out how to make a proper installer for windows: hint zipping up a lot of files is not the way to go. Windows users are used to easy downloads and then a double click to launch the installer. Technically easy to realize and yet Sun does not understand they need to fix this. At least it is multiuser now (that was downright embarrassing in 1.x).

Conclusion. Excellent if you are a casual office user and don’t need/understand word anyway. It’s has quite a bit of usability problems but seems reasonably fast and feature complete. Don’t expect pixel perfect compatibility with office but you will be able to open most documents and edit them. For me the lack of proper crossreferences is the showstopper. I need this and OOO doesn’t have it.