Now you can select a fourth output style for fil.tero.us: hReview. This is a microformat which is used to add metadata to identify the different parts of a review.

fil.tero.us will consider a tag consisting of one through five stars (U+2605 black star) a rating, and the “notes” field will be used as the review summary. The hReview standard supports other rating systems, but this was used to make the code efficient.

Try it out!

After tinkering for what seems like an eternity with a DB interface, I decided to just remove the sensitive information from the bookmarks file before writing it to disk. Hopefully there should be no more such issues in the future.

The main menu items should now always be the same height (set all element heights to 100%).

The “Clear” button will no longer drop down to the next line, because both buttons are now floated.

By making the output functionality just a little smarter, the XSLT will no longer output title="" for bookmarks without notes.

Pngcrush compressed all my PNG files flawlessly, saving up to 45% disk space on some files. Interestingly, the biggest files (by area) were compressed the most.

Pngcrush is free, open source, and the resulting files look identical to the originals. Highly recommended!

XSLT is quite flexible, and well suited for XML handling. However, PHP has oodles of string functions, most notably regular expressions, and I hope to enable those as soon as this version goes stable.

Update 2006-11-18: Rollback of changes until remaining issues with the PHP filtering are resolved.

Update 2006-11-22: Finished the new fil.tero.us! Some changes:

• About 10% faster than the XSLT based search.
• Non-ASCII searches should work properly. Try looking for e.g. Ω.
• Less clutter around the form.
• Better extension abilities with PHP’s feature set.

Matteo Penzo at UX Matters has published an excellent article, showing eye tracking data for different styles of web forms. I’ve applied their empirical data (and promise of world domination), and I hope you find the result satisfactory.

Update: This change had the unintended effect of making the layout work fine in IE.

This is something any site should have, and although this is a small site, I’m sorry that I didn’t add it before.

The summary isn’t entirely politically correct, but it summarizes the fix well. If you want to use fil.tero.us without synchronizing (which sometimes takes too long anyway), you can just login with an empty password field.

The site structure has been changed, so that instead of using e.g. “www.l0b0.net/blog”, the new link is “blog.l0b0.net”. The same goes for most of the other pages, such as news and fil.tero.us. The old links will continue working, so you won’t have to change any bookmarks.

Oh, and WWW is deprecated.

IE 6.0 couldn’t render the pages for several reasons:

• IE doesn’t understand the MIME media type “application/xhtml+xml”, which should be used for XHTML documents. When using it, IE treats the web page as a document to be downloaded.
• To force IE into treating XHTML as HTML, the media type “text/html” is used. This is not done for other browsers.
• Sablotron 1.0 (the XSLT processor used) uses the minimized form (<script ... />) when outputting “script” elements using the “xhtml” output method, even though the XHTML 1.0 standard says non-empty elements should not use the minimized form.
• HTML does not allow “script” elements to use the minimized form.
• For some reason, IE doesn’t fail to render XHTML documents as long as there is at least one “script” element which has contents.

The result was that IE choked completely on the pages when I removed the only remaining “script” element with some contents. Fortunately, I found a small hack which fixed the problem.

The previous version of these pages used to include a CDATA section in the header for the initialization JavaScript. This method resulted in uglier source, and no caching of this static information. So I’ve moved the script into a separate file. The markup is still valid, but this was the final drop to make Microsoft Internet Explorer 6.0 completely give up rendering the pages. More about this later.

del.icio.us announced today that they are now supporting https in the whole API. To improve the security for the users here, I’ve changed the relevant code to use the new URLs.

The "error handling" until now has mostly consisted of reloading until del.icio.us answers, or fiddling around the problems. Finally, you should no longer see any empty pages unless a fatal error is encountered. Errors will also be logged, so I should have a better chance of debugging problems. The log looks like ****, but at least it’s getting somewhere. Hope this helps!

An easier name for these dangerous times. Seriously, though, the name was a bit of a crutch, and most other del.icio.us tools have found catchy names using the del.icio.us theme. Which brings us right up to today :)

The new ISP doesn’t support symlinks, so I had to implement something in PHP for the users of the del.icio.us illogical interface to be able to grab the code. A little on the technical side, I first tried using highlight_file, but it was disabled by the ISP (php.ini directive “disable_functions”). The ISP allows overriding php.ini, but I’d have to override the whole file, which would require continuous maintenance, which is not very future-proof. highlight_string works essentially the same way, and it wasn’t disabled, so I went for that instead.

Earlier, if the del.icio.us service didn’t reply to the queries for synchronizing the bookmarks, the user would just see an XHTML error message in Firefox, since the native PHP error reporting function would just spit out some invalid XHTML in the middle of the page. After following the example in the PHP manual for how to set up my own error reporting, there should be no more of these problems.

After some annoying problems with the stability and configurability of my Blogger™ blog, I decided to start using Wordpress on my own site instead. I haven’t regretted it so far.

Because of some security related changes in the PHP configuration at the university, I have been forced to move the site here to be able to use the del.icio.us illogical interface. There are still issues to be resolved to make all the pages work here, but I hope to have the time to fix them in the next few days.