Thursday, February 7, 2002

AuthSample: Apple's preferred method of launching tools with superuser permissions on Mac OS X. It raises some good points, but also demonstrates extremely user-hostile behavior by installing a setuid root tool inside the application package, meaning that copying the application would give you a choice between not working or breaking the functionality. In a comment in the code they suggest that you prompt the user for reinstallation at this point. Hello?! I don't want to have to reinstall software because I decide to copy it around on my hard drive.! There are better ways. For example, the setuid root tool could be set up in a temporary directory. Or the application could include a non-setuid copy somewhere else, and reinstall it within its app package if it is missing. Security is important, but so is usability. One should not necessarily come at the expense of the other. 7:45:44 PM | reply [] I've been in e-mail limbo for a few days now, checking mail on my pair.com account because the machine on my desk I usually read mail from is having NFS problems and won't let me log in. I haven't been at school enough to get someone to look at it, so I'm compiling Mutt and Fetchmail (and their dependencies) on a Sun machine to use temporarily. It's taken an hour so far. This whole compilation thing is too freakin' hard and complex. Thank goodness for Encap, a university-developed, minimally invasive packaging tool that I use to keep my own installed software organized, but I'd like to be able to get source management too. So if I've got Mutt installed for IRIX, I can just point to it and say "I want that for Solaris too". 6:46:37 PM | reply [] This article does a lot to clear up the hype of the .NET CLR for me. When people have asked me for my opinion, I've expressed generally guarded optimism. Now I can be a bit more blunt and have the facts to back it up. I haven't read much about SmallScript recently as I don't write much Smalltalk any more; it's very interesting to hear about its limitations (and I'm looking forward to an AOS VM port for Mac OS X, which was promised at some point; SmalltalkAgents used to be Mac OS only). 5:41:15 PM | reply [] Wow, I knew NAI were morons, but it turns out it's worse than I thought. Please, please release PGP to someone who has a clue! 5:18:29 PM | reply [] Interview with KDE developer Cornelius Schumacher. "Q. You are hired to write the script for a commercial for KDE (like "The Heist" by IBM). What would be the plot? A. A big city at night, lots of lights. A thunder storm is over the city and suddenly electricity goes down, the city gets dark. But in some windows you still see some light from computer monitors. The camera zooms into the rooms and you see happy people using KDE. The slogan appears: "KDE - powered by community". Right on. Community—and idealism, in small doses—are wonderful things. (Now could someone port KDE to OS X? ) 4:51:35 PM | reply [] Got URL launching with ICeCoffEE working in Terminal.app, but it crashes at times. I think the problem is that no more characters are stored than there are characters in a line of text, so if I try to select from character 70, but the line only goes up to character 30, it starts merrily romping through memory. Unfortunately the structure that stores the Terminal buffer (_LineChunk) is not exposed at all, and I'm not very good at using gdb to figure out how it's structured. I've been poking at this for several hours, and I've reached the point of diminishing returns, so it's time to stop. What I may end up having to do is grabbing more text than I want, to the ends of the lines, and then postprocessing the result. This is only going to work in the Mac OS X 10.1 Terminal, it was significantly rewritten from the Mac OS X 10.0.x Terminal: for example, the class that shows the terminal view is called "TermView" in 10.1, versus "Terminal" (a subclass of "FieldView") in 10.0. 6:27:30 AM | reply []