Le lun 29/09/2003 à 23:38, Seth Nickell a écrit :
> It's nice but it won't fly with java daemons for
example.
> Java people do not care nor want to know about system stuff. You'll
They do on "that other platform". Anyone who wants their code to be used
either has to find somebody to do the "last mile" work for them
(distributions currently)
Or community packaging projects:).
But I freely admit packaging java is a major pain - most upstream
developers have not got the faintest idea what a properly managed system
can look like (like they do not even know they have a problem).
Some java apps are really nice once packaged though.
, or they have to do it themselves.
Nice dream:)
IMO,
distros got into this "last mile initscripts" business because their
systems were different and incompatible, not because it was usually a
good thing.
As someone that has written its share of last mile scripts I'd very much
like not to have to dump them now.
> always need a shell wrapper to launch the jvm for example (well
people
> may use gcj one day but you get the point), change system users and so
> on.
Well, until we get dbus/shell bindings, you can't write shell scripts
that aren't "legacy initscripts". But you can write tiny python
wrappers, and probably perl in the not so distant future.
Well if one has to do python or perl you'll make it that more difficult
to package non-python or non-perl stuff. When the projects we work with
provide an unix layer it's always in shell (and most often in a very
primitive dialect since Jakarta for example wants its scripts to run on
everything from cygwin to os X including strange and retarded Sun/HP
systems).
SystemServices
probably makes it much easier to write shell wrappers than initscripts,
just because the necessary boilerplate is like 4x less.
I'll wait eagerly for the shell interface then.
Cheers,
--
Nicolas Mailhot