"flow"/"recipe" as a basic ARK thing?
Will Partain <[email protected]> Mon, 05 Aug 2002 15:44:55 +0100
| Newsgroups | gmane.comp.sysutils.ark.devel |
|---|---|
| Message-ID | <[email protected]> |
Folks, just an idle musing that (I hope) is not a million miles from stuff Jonathan was writing about last week (hairy XML/Schemas/XSLT/... fun). When you buy into doing package stuff in the Sidai Way, the thing you're really buying is a particular "flow": beginning with <get-source>, on through <configure> and <compile>, to <install> and then finally to <deploy> and <reveal>. All packages then go through that "flow", even if some of the methods are made to be no-ops. The GNU package flow is a specialization of the Sidai flow: basically, it has a <configure> step that does something. You might have "flows" for "add a new user", "rotate all log files at a site", and so on. Can we say more powerful/stronger things about a "flow" than we might about a single method? Can a flow be a more bolted-down thing, so that the person re-using it is more confident of its behavior (e.g. a "flow" might specify all possible <param>s, so that a specializer couldn't add any weird and wacky code to it)? Are flows kinda orthogonal, say something like "aspects" are to object-oriented programming? I'm wondering if we have a candidate for a better first-class abstraction... (I'm mumbling now; I will be mostly off the air for a couple of weeks.) Will ------------------------------------------------------- This sf.net email is sponsored by:ThinkGeek Welcome to geek heaven. http://thinkgeek.com/sf