I’ve had a notional wish list for forum software for ages. Right at the top of that list is “organise using categories rather than separate forums”. Just as I was about to get off my arse and do something about it (well, nearly) I stumbled across Vanilla. It does that and has a bunch of other features I like too. Lots of nice little AJAX (urk) touches in there as well.

If you’d like to come and play I’ve set up a Hexten forum here. I haven’t done much to customise it yet. Apart from turning on a few extensions that’s how it comes out of the box.

I’ve been playing with Google Analytics. It promises some pretty sophisticated looking server log analysis. I’m a sucker for web logs - tail -f access_log is one of my favourite TV shows - so I’m always excited about new log analysis tools. Unfortunately I haven’t yet experienced the thrill that only a tidy looking pie chart can provide with Google Analytics. I’ve added five sites to my profile and placed the required tracking code on them; I’ve verified that Google sees the tracking code and yet, some 30 or so hours since I added the first site Analytics can’t draw me a graph.

Admittedly the reason why I didn’t see any stats in the first 20 hours was because I was using Safari. Eventually I found a well hidden fragment of help text that explains that the reports don’t work with Safari. Cool that a service which identifies users’ browsers as one of its functions can’t identify my browser and tell me it’s not supported.

Switch to Firefox and I get a little further - now I get the graphs but they’re all saying there have been no hits at all on my site. Maybe I’ve offended somebody?

Bear in mind that this isn’t the usual perpetual Google beta; this thing is supposed to be ready for prime time. I suppose it’ll work eventually but it certainly doesn’t work as advertised. One thing I do like: even though they’re not supporting the Mac’s native browser they have paid a back handed complement to Cupertino by ripping off OS X’s little progress spinners. So that’s nice.

Update: Peter Strand has instructions for configuring Firefox so you won’t be tracked by Google Analytics here. Slightly ironic then that he’s using Google Ads (essentially the same technology) on his page.

Another update: This just in from Google Analytics Support:

From: "Analytics Support"
To: andy
Subject: Re: Tracking has not been validated or added…

Hello,

Thank you for your email and your patience.

We have received your report regarding the problem with the "Check Status" alert update. Our engineers are currently working to solve the problem and hope to reach a resolution shortly. This will not affect data collection or report generation if you have already tagged your website with the Google Analytics Tracking Code.

Additionally, I understand that you aren't seeing data in your reports, even though your tracking code has been set up for over six hours. We apologize for any inconvenience. We have collected your data since you installed the tracking code on your site, and are continuing to collect this data throughout the day.

We are currently in the process of creating reports from your data. You should be able to see your reports populated with data later today. Please note that this reporting delay is associated with unexpected demand for Google Analytics. Under normal circumstances, the data in your reports will be at most six hours old.

For additional questions, please visit the Analytics Help Center at http://www.google.com/support/analytics. If you're unable to find an answer to your question on our site, please feel free to reply to this email.

Sincerely,
Analytics Support

Summary: we’ve been swamped by the demand. From poking around in the tracking Javascript and information that’s revealed by the error messages I’ve seen I’d guess that they haven’t yet migrated the service to their legendary server farm. Maybe it doesn’t scale that well yet? You’d have thought Google - of all people - would have braced themselves for the traffic, no?

This is one of the best bits of software news I’ve heard in a while - and pretty much entirely unexpected. Xara Xtreme is going open source with Linux and Mac versions licensed under the GPL.

If you haven’t heard of Xara it’s one of the finest vector drawing programs ever. Its roots are in a product called Artworks which ran on Acorn RISC OS machines. The first version of Xara I owned ran snappily on a 486, did real time anti-aliasing and basically knocked spots off anything else. My venerable copy of Xara X is the main reason I boot up a Windows machine these days.

Xara (the company) started out as Computer Concepts and they produced the first piece of software I ever bought - the WordWise wordprocessor for the BBC Micro. Legend has it that Charles Moir wrote it at his mum’s kitchen table. About fifteen years ago my company and Computer Concepts briefly collaborated on a hardware product and, frankly, I can’t wait to have the chance to work with these guys again - they’re an excellent company who have been producing top quality software for twenty five years.

I’m rambling now… Just so excited that not only will I be able to run Xara on my Mac and Linux boxes - I’ll probably be able to help get it there.

Working on K8055Mac has got me thinking. A few years ago I worked on Total Control which was an educational product for Acorn RISC OS computers. It came with dedicated (and relatively expensive) hardware and was designed to be used in the teaching of control theory in UK schools. Changes in UK educational policy and our tardiness getting the software out meant that it never really found a market - although the few people who trialled it loved it.

I’m just wondering if there’s any interest in an updated version running on Mac OS… Mainly of interest to educationalists I imagine.

A couple more screenshots…

I’ve been thinking about how to automate the process of making a new release of a project I’m working on. Ideally when I decide to make a release I’d like to test the build and install on all the platforms I’m supporting. The best way I can think of to do that is to have a bunch of virtual Linux (VMWare, Xen or whatever) boxes which I can boot up, build and install the project on, log the results and then shut down.

I’d like any changes written to the filesystem of the virtual box to be transient - so the next time I use it to test the build it starts up in the same known state. The obvious way to do that would be to have a block device driver that starts out just like the loopback device (i.e. reading from a file that represents a block device) but have it capture any writes and store any written data in a separate data structure - a delta.

That would mean that at any time you could roll the filesystem back to its initial state by discarding the delta. Alternately you could tell it to merge the delta back into the main filesystem and thereby commit any changes you’d made.

Thinking about it a bit more it seems that this would have all sorts applications. Rolling back installations that have spewed files all over your disk; safely finding out what a virus does; the ability to roll the filesystem back if you screw something up yourself (hands up who’s never done that…).

If it was fast enough you could run your main filesystem in this mode and have a second partition or device to hold the deltas. Every now and then you’d commit the delta to the main filesystem and start again.

If the delta was organised as a log file you could choose how much of it to commit - or roll it back to an arbitrary time - “OK, let’s go back to 9:36 before I screwed everything up”. Having all the writes available in chronological order would also make it possible to safely commit the contents of the delta without unmounting the filesystem.

Now I’m assuming this is so useful that it must already exist but I can’t make Google tell me about it. SEH isn’t quite what I have in mind; I guess this paper describes something like it. If it doesn’t exist I think it should…

The Cherry CyMotion Linux is a high quality keyboard with Linux specific keycaps. It’s supplied with a version of SuSE and a kernel patch which didn’t work for me against the 2.6.11.7 kernel. I’ve manually applied the changes it made (updated to use the idioms current in 2.6.11.7). Here’s the patch: cymotion-2.6.11.7.patch. Let me know how you get on with it.

2005/08/26 Update: It turns out that, at least on the PC I’m using, the CyMotion isn’t recognised by the BIOS although it works fine once Linux has started. That seems a bit sucky… Other USB keyboards work fine. Is it just me?

2005/09/06 Update: It was just me… Didn’t have the BIOS set up to identify USB devices. Silly me.

Sick of brute force password recovery attacks on your server and maybe a little worried that one of them might actually succeed? You need pam_abl. It’s a little PAM module that automatically blacklists hosts that are responsible for large numbers of failed authentication attempts.

I’ve just released mod_log_rotate which provides server-native log rotation for Apache 1.3 and Apache 2. More about it here.