I really think the scripts are more flexible and powerful than command metadata. But I agree having to ship them is a pain, so I hope we can find a better way in the future. Your proposal of a known location in the bundle makes sense.
2016-10-19 13:51 GMT+02:00 Christian Schneider <ch...@die-schneider.net>: > The command metadata service API of course would need to live at felix > gogo. So the plain gogo shell could also make use of it. > We could then provide an extender at karaf that creates these services out > of the karaf specific commands. > > The problem I see with the scripts is how to add them to the karaf shell > in a modular way. After all you want to just install the bundle with the > command. > You do not want to additionally have to install a shell script at some > other place. That is where the magic location for a script file could help > for the short run. > > Christian > > On 19.10.2016 13:43, Guillaume Nodet wrote: > >> I thought about that, but then, the bundles would be tied to Karaf while >> the goal is to support commands from projects which are not related to >> karaf. Else, we could just force them to use our own command api if they >> want to support Karaf ;-) >> >> Also remember that scripting offers very nice additional features : it's >> dynamic. A static completion system would not be able to offer the same >> level, so you'd have to offer an api, and then... well, it's not >> independent anymore, so you've lost the benefit. >> >> The fact that the scripts are external it not necessarily a problem. >> That's the way your unix shell completion works btw ;-) >> >> > -- > Christian Schneider > http://www.liquid-reality.de > > Open Source Architect > http://www.talend.com > > -- ------------------------ Guillaume Nodet ------------------------ Red Hat, Open Source Integration Email: gno...@redhat.com Web: http://fusesource.com Blog: http://gnodet.blogspot.com/
