I should have taken a real vacation back in March instead of the half-assed attempt I made. In the hope that it's not too late, I am taking that vacation now.
Emacs-w3 does not seem to have built-in support for gunzipping stuff automatically. It calls out to "gunzip -c" to handle it. Sometime I'll look up the Emacs-lisp magic to call that on a buffer and replace the buffer contents with the output. In the meantime, I've taken the cheap way out and disabled the sending of "Accept-encoding: gzip".
Posting works pretty much fine, except for a weird error on the return from the server (the post goes through, though). But retrieving entries is broken. I may take the time to figure out why later, but work calls now...
Update: Well, looks like I needed to tell emacs-w3 to not claim to support gzip... The XML-rpc code was not dealing well with parsing gzipped data. Next step—hook things up so gunzipping can happen properly
Today I found myself walking home from campus, mentally comparing PDF (editable) versions of tax forms to paper versions. Without exception, the paper ones take a lot less time to read and fill out. The instruction booklets are easily skimmable, unlike PDF. Taking in the whole form at a glance or focusing on a particular part is a matter of minute, instinctive hand movements (compare this to zooming PDFs). There's no comparison between readability of printed text and text on a monitor, of course.
All well and good, but then I got to thinking about other computer programs I've used, not just Acrobat Reader. With the single exception of console video games, I find all of them less usable (from the "easy things should be easy and hard things possible" point of view) than I find a Form 1040 plus instruction booklet.
That's a scary thought to anyone who's ever looked at the 1040 instruction booklet.
At first glance, the comparison is unfair—the paper 1040 relies on skills that it's taken me years (if not decades) to learn. But learning those skills (reading, writing, skimming for content) is something that has incredible breadth of application. The learning that has to happen to make computer programs more usable tends to only be applicable to that one program...
I still haven't figured out what it is about console games that tends to make them so much more intuitive and usable... It makes sense, from a business standpoint -- no one will play a game that's not a pleasure to play. Do game designers carefully analyze the user interaction? Perhaps they could share some tricks with other software authors...
In other news, Matt and I spent some time today talking about dimension theory, slowly working though the proof of Krull's Hauptidealsatz. Things are making more sense, slowly... very slowly.
I feel compelled to add my useless opinion to the Firebird naming controversy. It's a 3-part deal:
Over the last week or two, I've written up a little document on the workings of the URILoader. If you read it and have comments, please let me know—feedback is much appreciated.
Once I've gotten some comments from Darin and Seth Spitzer, I'll likely check this in; at that point the URI will change (so don't bother to bookmark it). I'll update the link in this entry when the document moves, though.
Update: the document has a home on mozilla.org.