2014-09-23 10:34 GMT+02:00 Felix Meschberger <[email protected]>:

> Hi Chetan
>
> Am 22.09.2014 um 19:05 schrieb Chetan Mehrotra <[email protected]
> >:
>
> > On Mon, Sep 22, 2014 at 10:28 PM, Carsten Ziegeler <[email protected]>
> wrote:
> >> Right, but there is nothing specific we have to do in the model about
> it.
> >> You need a way to inject this stuff on deployment, and this can be done
> >> with merging a deployment model into this model. And I guess there are
> >> other ways to do it.
> >
> > Not sure what those other ways should be. For me the sling start model
> > expressed via some xml/json/yaml etc seemed to be like a single and
> > sufficient information source to launch a working application. If
> > property interpolation is not supported and I need to use same model
> > to be used to setup Sling for dev/test/staging/prod env then I am not
> > sure how environment specific configuration would be managed and
> > tweaked without duplicating the whole model for each case
> >
> > Perhaps the design is in early phase I should wait for some real world
> > examples on how this is supposed to be use.
>
> Ah, this is the issue: The Sling start stuff is *not* for starting an
> instance but for describing the build of a binary which can subsequently be
> used to launch instances.
>
> Yet, you indeed hint at an issue I am currently having will  the
> crankstart and slinglaunch and whether currently is being though of:
> Everything scratches a different itch around the same problem group of
> describing a binary and setup different launches.
>
>
The model is not necessarily describing a binary containing all the
artifacts. It describes a deliverable or installation. How this is started
is a different thing. So crankstart can be used to process the model
pulling the artifacts from a repository, launching the framework installing
the bundles etc. Launchpad can be used to create a binary based on the same
model.

Carsten

-- 
Carsten Ziegeler
Adobe Research Switzerland
[email protected]

Reply via email to