Alan Cabrera wrote:
My humble thoughts:
Like any good work of art it is hard to know when enough is enough and
it's time to leave the piece as it is. I am worried that what was
once a clean and elegant spec will become muddied to support features
that, imo, tools should have taken the burden; take required bundles
for example.
I couldn't agree more. However...
I'm not aware of the compelling use cases that call for an extension
to the number of hosts for a fragment but can think of one, again imo,
not so compelling one. This being the patching of a set of bundles
with one fragment.
For me, the issue with fragments is not the concept, but the realization
of the concept which turned out to be a little ugly. The main use case
for fragments was localization, since such info needs to be in the same
package as the class being localized. The realization extended beyond
this use case. With respect to the issue at hand, I don't know enough
about localization to say if it is compelling to have one fragment
attach to multiple hosts, but I could imagine a scenario where a suite
of bundles somehow package their localizations into shared fragments so
that they don't all have to maintain their own copy.
Required bundles are a whole other can of worms...
I'm happy to go into detail as to why I think poorly of this use case
if it is the impetus for adding this extension and if this is the
appropriate forum.
Well, I don't think this mailing list is reserved purely for praise of
the spec, but I guess there isn't much use in crying over spilled milk
either. :-)
-> richard
About PackageAdmin.getHosts(), I assume that for the moment it should
always return a single host. Just curious, was there a reason to have
it still return an array and not a scalar in light of the current
cardinality between host and fragment?
Is there a way for me to know what's in the pipeline so that I may put
in my 2 cents?
Regards,
Alan
On Oct 2, 2007, at 5:42 AM, BJ Hargrave wrote:
:-)
During R4 spec development we allowed fragments to attach to multiple
hosts. Prior to going final on R4, we changed course and limited
fragments
to only attach to a single host. The sentence you quote is left over
from
when we considered allowing fragments to attach to multiple hosts. Also
see the method signature of PackageAdmin.getHosts where the return
type is
an array and not a scalar.
We are currently discussing revisiting that decision. That is, to
potentially allow a fragment to attach to multiple hosts. It would be
useful to get input from the OSGi community on the pros and cons of
such a
change.
--
BJ Hargrave
Senior Technical Staff Member, IBM
OSGi Fellow and CTO of the OSGi Alliance
[EMAIL PROTECTED]
office: +1 386 848 1781
mobile: +1 386 848 3788
From:
Alan Cabrera <[EMAIL PROTECTED]>
To:
OSGi Developer Mail List <[email protected]>
Date:
2007-10-02 02:07
Subject:
[osgi-dev] Finding Localization Entries
Hello,
I'm trying to get my head around 3.10.1 Finding Localization
Entries. The last sentence of the first paragraph confuses me.
"Fragment bundles must delegate the search for a localization entry
to their host bundle with the lowest bundle ID."
Does this mean that fragment bundles can have multiple hosts?
Regards,
Alan
_______________________________________________
OSGi Developer Mail List
[email protected]
http://www2.osgi.org/mailman/listinfo/osgi-dev
_______________________________________________
OSGi Developer Mail List
[email protected]
http://www2.osgi.org/mailman/listinfo/osgi-dev
_______________________________________________
OSGi Developer Mail List
[email protected]
http://www2.osgi.org/mailman/listinfo/osgi-dev
_______________________________________________
OSGi Developer Mail List
[email protected]
http://www2.osgi.org/mailman/listinfo/osgi-dev
