MishMash

UI is Better than AI

In an article about the differences between desktop and server based RSS aggregators, William Grosso coins the phrase

UI is Better than AI.

which is a pleasingly catchy encapsulation of an idea which I was thinking about during my first weblogging experiment, and fits with a theme which I have returned to recently. The first of the above linked articles includes two quotes from a paper in the Journal of Hydraulic Research (the IAHR web site is a royal PITA, with Frames-and-Flash based navigation which doesn't work in Konqueror, and doesn't allow deep linking without picking frames out of context). Since that post is on a weblog which in theory could disappear at any time, I repeat them here.

Knowledge based systems may support certain tasks, but if the task is restructured, the need of much of the knowledge about the complexity might become obsolete. Much of the knowledge which is still needed now is simply an artefact of certain solution methods or of the use of certain computer programs. Modification of the solution process and improvement of the tools tend to make the task less knowledge-intensive. [p. 91, van Zuylen, Dee, Mynett, Rodenhuis, Moll, Ogink Most, Gerritsen, Verboom (1994) "Hydroinformatics at Delft Hydraulics", J. Hydraulic Research, Extra Issue: Hydroinformatics]

…

Tools were proposed to provide knowledge in a suitable form ... Once it became clear what kind of tools could be implemented, the developer and the expert/user restructured the task in such a way that it became easier, less complex and less knowledge-intensive. A user interface was then developed for this task. Clearly, a straightforward development of a user interface containing a knowledge base would have resulted in a sub-optimal product, since the product would have supported an obsolete task. [p. 93, van Zuylen, Dee, Mynett, Rodenhuis, Moll, Ogink Most, Gerritsen, Verboom (1994) "Hydroinformatics at Delft Hydraulics", J. Hydraulic Research, Extra Issue: Hydroinformatics]

Server nightmares

Geek-speak warning.

Ouch, yesterday was painful. As I mentioned in the last post, the server which runs the FloodRiskNet web site and more besides died. I spent almost all day (from morning coffee time when I rebooted the server until eight in the evening when it next booted successfully from its hard drive) panicking, running up and down stairs with various rescue CDs in the hope that one of them would provide me with some useful tools for tracing and fixing the problem.

The symptom: The machine would start, the boot loader would run, the kernel would load, and then it would freeze just after printing

Freeing unused kernel memory: 130Kb

With the help of Google, I found out that this is the point at which init starts. Init is the first process started on a Unix box once the kernel is running. It took a while from that discovery until the penny dropped that this probably meant that there was an incompatibilty between the kernel, which was old, and init, which was relatively new because I was keeping the system software up to date.

The kernel was old because I am using the XFS filesystem, and have only recently gotten round to working out how to build patched kernel packages in Debian. Since the stock kernels do not include XFS support, I couldn't upgrade the kernel along with the rest of the system using the standard kernel packages, and I was always intending to work out how to build one but not getting round to it.

I might have found out earlier that there was a problem but for the fact that it isn't necessary to reboot Linux for much. It continued running quite happily through several upgrades, but, I presume, although init was upgraded it was never restarted, so the new version wasn't tested. It was therefore only when I rebooted that the problem manifested itself.

Once I'd tracked down the cause of the problem, the fix was relatively easy, although it necessitated a bit more learning of the workings of GNU/Linux. In particular I was very glad that I'd heard of chroot some time, because if I hadn't I still wouldn't have a working server. With it I was able to boot using a rescue CD, mount the hard drives from the system in the correct configuration, and then chroot to the system's normal root directory. That allowed me to use the system package management facilities to install the new kernel package that I'd prepared (copied over on a 1680Kb formatted floppy — that took a while too, because I only have old and dodgy floppies lying around).

That done, it booted right up no problem. And I did a little dance, which was fine because it was 8 o'clock and no one was there to see me.

The moral of the story is probably don't put things like kernel maintenance off. Hmmm.

A twist in the tale is that I forgot that shutting down a system booted from a rescue disk wouldn't unmount any filesystems which had been manually mounted. This meant that at some point the system XFS filesystems were unceremoniously turned off. Now XFS is a journaling filesystem. The whole point of using it was reliability; if the system is turned off without unmounting the filesystems, then there is a journal on the disk which can be replayed. There is a much lower chance of corruption. Rather upsettingly, I ended up having to run xfs_repair and force it to fix one of the filesystems because it was damaged in a way which meant it wouldn't mount (and therefore couldn't replay the log). Precisely the situation I was trying to avoid by using XFS.

Recent travels: Delft, Brussels, and Nottingham

It's been a busy couple of weeks, culminating with a panic yesterday when I thought I'd broken the FloodRiskNet server, which is now running a few other sites as well. I wanted to run a backup before opening the case to upgrade it, but decided (don't ask) to reboot it before do so. Of course it failed to boot. I ended up having to work out how to upgrade the kernel on a (Debian) box I couldn't boot from the hard drive. A little more knowledge accumulated …

I've given three presentations in that time, the first (already mentioned) was OK, the second better (Slides: HTML, PDF, 290Kb), and the third not so good again (my impression, though I'm not necessarily the best person to judge). The reason for the dip at the end was mostly that I was conscious of time, and hadn't had time to optimise the slides for available time or for the audience. That was a shame, since the audience was our project partners to whom we hadn't previously given a full explanation of what we are doing. On the other hand since they are our colleagues we will have further opportunities to expand on what I said.

That was on Monday, having arrived back from foreign parts on the Saturday night. I've started at the end already, so what the hell, I might as well show my true colours as a chronological pervert. I spent Friday night with Owen in Brussels. Owen is Tom's brother. Tom is an old friend, and I met Owen I think for the second time a few weeks ago in London while I was over visiting Tom. I forget whether Owen invited me then or whether I invited myself; either way I took the opportunity to have him show me some of his Brussels, and if I wasn't invited, he did a good job of hiding it. I was stopping in Brussels anyway (more about that later, or rather earlier) and Owen had declared himself to like the place, so it seemed like an ideal opportunity to get to know both Owen and Brussels a little better. In the end I didn't learn a great deal about Brussels — it has at least one decent bar, and at least one dreadful night club — but I did have a fun night with friendly people.

So thanks, Owen, for your hospitality.

I was in Brussels to visit Mike Abbott and his wife Louise. I was expecting to meet Mike in Delft, but Louise generously suggested that I visit them.

Mike is even more enamoured of Brussels than Owen, extending his enthusiasm to cover all (or at least most) things Belgian, and made an excellent tour guide. There are some spectacular palaces in Brussels, whole streets of them. Many now seem to house designer clothes shops, though it is apparently possible to pick one up for next to nothing (the maintenance costs are terrible). They tend to have a front window, and a passage running through from front to back. The entrance lies half way back along that passage. A perfect layout: the owner of the house can alight from their carriage under cover and right onto the door step. In stark contrast to the UK, there is no sign of entire ground floor fronts being removed to make way for side to side, top to bottom plate glass and a tacky plastic sign. Impressive.

Thank you Mike and Louise (and Picasso, Matisse, and Cindarella, the cats who run the show) for a warm welcome.

I travelled with Mike to Brussels from Delft, where I spent a few days for a number of reasons. He had some time ago made me and Andreja Jonoski aware of each other, and encouraged us to make contact, and I had originally thought to travel to Delft primarily for that purpose. The department of Hydroinformatics and Knowledge Management at the Institute of Hydraulic Engineering (now UNESCO-IHE) hosts a number of people doing interesting work who I was keen to meet in any case.

Since it seemed likely to be of interest there and to generate useful feedback, it seemed an excellent opportunity to take my current research work out for a public airing, hence giving the presentation.

As I was building up to arranging this trip, its nature was changed somewhat by the fact that the department advertised a vacancy for a lecturer in Hydroinformatics Systems. As a result I spent some of my time there discussing this with the staff of the department.

I also took the opportunity to catch up with Peter Gijsbers and Stef Hummel, both working at Delft Hydraulics on the HarmonIT project. It turns out that they are asking similar questions in that context to those I am exploring in my work, which led to a productive conversation.

All in all then I was made to feel welcome, I learned a lot, and the work I presented generated some challenging questions and some positive feedback. Whether or not I move to Delft at any point, my interests are likely to continue to overlap with those of many people in Delft, and it is good to have a picture of them in their environment; it provides context and lends a more personal feel to the impersonal medium of email.

Thanks then, finally, to all at IHE and Delft Hydraulics who made the trip interesting, fun, and valuable.