felix-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Carsten Ziegeler <cziege...@apache.org>
Subject Re: [DS] Various factory component issues, and progress on RFC 190
Date Tue, 20 May 2014 05:39:04 GMT

2014-05-19 22:57 GMT+02:00 David Jencks <david_jencks@yahoo.com>:

> There are some difficulties with factory components and DTO/mutliple pid
> support.  In terms of the spec, DTO and factory components don't work well
> together, see bug 2683 (I'm not sure how visible this is to
> non-osgi-members).  In Felix,
> 1. I think we should not support the obsolete factory configuration >>
> newInstance-like behavior on namespace 1.3+ components.  We shouldn't have
> supported it after namespace 1.0, but we should stop now for sure. Trying
> to support this with multiple pids is just too weird.
> +1, yes we should have dropped that support with earlier namespaces :(

> 2. Because of how I implemented multiple pid support for normal
> components, I think it would be easy to Implement factory config > multiple
> instances of the ComponentFactory service.  I believe this behavior was
> requested by Pierre de Rop.  There's an additional problem if you do this
> that you can't distinguish the ComponentFactory services registered from
> the multiple factory configurations.  To solve this I propose that
> configuration properties prefixed with "org.apache.felix.scr.factory."
> (exact string up for discussion) be added to the ComponentFactory service
> properties.  Of course all this would have to be enabled by an flag in the
> xml such as felix:factoryComponentFactoryPID="true" (no good ideas on this
> one yet).

Sounds fine to me. Would it make sense to get this into RFC190?

> As a reminder, a few days ago I said I was about to remove the legacy
> configuration system entirely in favor of the spec one based on DTOs since
> they are conceptually incompatible and the existing one is fundamentally
> confused.  So far no response so it's going soon….
I agree, let's just go forward with the DTOs


> Comments?
> thanks
> david jencks

Carsten Ziegeler

  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message