On Saturday 21 April 2007 09:15, Aggelos Mpimpoudis wrote:
> Would you encourage a new OSGi developer, to try SAT from his first steps?
> Is it safe to dump the servicetracker and switch to SAT method at this
> early steps or would you suggest to keep on with the first?
There are a few facets to a question like this;
1. By adopting such utility (I don't know if this particular one, but there
are many others) you might buy some coding time initially, and it may be
the tie-breaker to go ahead or not.
2. It is likely that such systems are more aware of the tracking issues and
could be a lot less buggy than hand-written trackers. Also, if the
community around the project is strong, it is likely that any problems
can be fixed by 'others'.
3. I have done quite a bit of mentoring on OSGi projects, and most have tried
to move away from the manual trackers. However, that means that you end
up further from OSGi, and placing a 'learning curve' on the mentor, ending
up spending more on the mentorship.
Later on, this also applies to hiring new developers. Their previous OSGi
experience is diminished some what.
4. Personally, I think most such systems gives a short-term gain and
long-term pain. I typically recommend that the trackers belong in the
bundles that registers the services, and the 'client code' instead just
do a;
MyService service = new MyTrackingServiceImpl( bundleContext, ... );
i.e. the tracker implements the service interface and contains the
default fallback strategy.
Now, that said; your own circumstances are of course unique, and general
advice is not always the best for you. If the project is related to teaching,
I suggest to stay "close to the metal" and not hide too much.
Cheers
Niclas
_______________________________________________
OSGi Developer Mail List
[email protected]
http://www2.osgi.org/mailman/listinfo/osgi-dev
