On Mon, 2005-08-29 at 19:34, Tim Nelson wrote: ... > He's adding stuff to the end of the filename. So, for example, > say we want file foo on servers that are both mail servers and web > servers, we get this: > > Mark (msa) solution: > .../usr/local/bin/foo.__MailServer.WebServer__ > > Tim templating solution: > .../MailServer.WebServer/usr/local/bin/foo > > Naturally I like mine better, but his solution should work just as > well if that's the way he wanted to do it
I have no problem with Mark doing it "the way he wants to do it", but these aren't at all equivalent solutions to the question of how to organize your sources, and he was asking for feedback about his approach. The first obliges you to individually copy the files (or use some secondary renaming mechanism), the second allows you to recursively copy a whole directory tree for 'MailServer' or 'solaris' or 'foo_domain_com' or 'DevelopmentNetwork' or whatever. The first would quickly lead to some complicated naming conventions, and equally complicated filters and includes/excludes. My point, and my opinion, is that a filesystem-mirror tree as an organizing principle for all of your source files isn't particularly useful or scalable. Mark asked 'Are you actually doing this?', meaning copying from source directories related to classes. I'd guess that most EVERYBODY reading this list is actually doing that. I'd be more interested to know if anybody is 'actually' copying from a completely filesystem-based repository tree structure. -Ed _______________________________________________ Help-cfengine mailing list Help-cfengine@gnu.org http://lists.gnu.org/mailman/listinfo/help-cfengine
