Human-readable data serialization language for Perl, Python, Ruby, Java, etc.

Ogg Vorbis listening samples side by side with WMA, MP3, RA, et al.

Compound Text is a format for multiple character set data, such as multi-lingual text Images missing, get the PDF from http://www.x.org/X11R6.8.1/docs/CTEXT/ctext.pdf or the sources from http://cvs.freedesktop.org/xorg/xc/doc/specs/CTEXT/ctext.tbl.ms?rev=HEAD&view=markup

Just for a quick review, here is the dump I got by scraping the site the other night (and/or day) This is temporary, and will disappear. Soon. It's just for you to comment on before I do the next round (tentatively, I'd do this weekly). Please review the details of the YAML formatting in particular, if you think you know how it should look. *** This is a 770,167-byte file on a slow server [now up to 1,3 Mb]. Don't click if you're impatient. I deleted entry #247 altogether. Other than that, the data is as I got it off the server. I obtained the data by going through the /entry/nnn for each nnn up to number 2519 [now 4259], which for some reason I believe was the highest entry number at the time. Then I extracted the actual data out of the web page for each entry, and converted it to (some sort of) YAML. The /entry/ page doesn't show the modification time, so I don't have that. I have reserved a field for it in the format nevertheless. I also have this in something resembling XML, of course without any formal DTD or anything. Let me know if you think getting that would be more useful (instead, or as well). For the context impaired, the "and/or day" refers to time zone confusion.

This (Windows!) program managed to salvage almost all of my trashed ext3 filesystem

The PU1 PU2 PU3 PUA and older Ling-Spam email corpora

This Monday's dump is now available, at the same location as last week The old data from last week is in entries-2005-04-07.yml. The file is now 1,343,896 bytes, and the server is still slow. I've scraped up through entry #4259, for a total of 4049 undeleted entries. (Hmm, I guess I'm falling behind pretty badly already.) And the location is still very temporary. Still, here's the URL again: http://www.iki.fi/era/tmp/entries.yml The scrape took several hours, and involves a separate TCP connection for each page fetched. I guess I ought to make that slightly less resource-intensive, but I'm not sure how exactly to do it. The current system was made that way because I wanted to be able to specify an arbitrary pause between fetches (currently 4 seconds). Will I be tying up resources if the fetch keeps the connection open but idle for several seconds between each fetch?

Troff source for ctext spec at x.org (via ViewCVS at Freedesktop.org)