Wes Felter pointed to a note on the advantages of the NT registry on the Linux-future mailing list.
The advantages given for the NT registry are typing (Is this value supposed to be a number or a string?) and fine grained security.
Unfortunately, while I can see the point of those advocating knowing all of this stuff about data, I'm beginning to wonder if this attitude isn't indicative of a larger problem with the computer world these days. On a user level, it's people sending me email messages as 100k Word documents which, when unpacked and transferred to a system I can read them on end up being 2k worth of "different types of poops" jokes.
Does 98k of font and formatting information really make feces any funnier?
I'll leave my rant on people whose web pages have blinking rollovers which lead to "under construction" pages for another day 'cause you're already sick of it.
On a more technical level we see API and design bloat. Do I need 5 different ways to read a file? How much easier is it to read polygons[4] than ListIndex(polygons, 4)? At first glance there's a difference (the Macintosh ransom note effect), but if this is something you're working with day in and day out?
Back to the question of the NT registry: Most Unix applications allow configuration on a global level, then overriding on a local level some or all of the configuration options via a local '.rc' file. The global file can belong to a number of groups that you can assign your higher level admin types to, and the local file is owned by the user that the little tinkering things affect. How does this differ functionally from the registry? Not a whole lot, but because the structure of the configuration is somehow more rigid, it's quite a bit more attractive to some people.
Oh to be young and impressed by academic "cleanness" rather than raw functionality again.
Thursday, February 11st, 1999 danlyke@flutterby.com