Hi, > There is a nice definition in > http://citeseer.comp.nus.edu.sg/676839.html > > The authors come to the conclusion that strong migration is impossible > in the JVM. I find their arguments very convincing: strong migration > means the program can be unaware of being migrated and involves sending > stack, instruction pointer and registers along with the code while weak > migration can be triggered by the program itself and only sends the code > itself. As you already mentioned, weak migration does not transport more > state of the program than is being actively transmitted by the program. > Yeah, I am greeting their arguments with an applause too. They are right that it is impossible to support strong migration in Java without breakign Java model or instrumenting byte code.
> A complication is that in a running OSGi system it may not be enough to > just transmit one bundle's state since that state might depend on other > bundles as well. Yes, I will say with you. I must to take into account that in OSGi platform, one service can be dependent from other service and their states are not known in priori. > This leads me to believe that weak migration is quite feasible in an > OSGi platform (just grab the bundle's bytes, e.g. using the > Bundle-Location: header, and send them elsewhere) while strong migration > is probably to hard to be useful. > >From the header of one bundle, can we determine that from which bundles the >regarded bundle depends from? I only know that there is a statement >Import-packages that declares which packages the bundle use. In other words, >how can we define those imported packages belong exactly to some other >specific bundles and it is possible to find those bundles? Kind regards, Conan. -- Ist Ihr Browser Vista-kompatibel? Jetzt die neuesten Browser-Versionen downloaden: http://www.gmx.net/de/go/browser _______________________________________________ OSGi Developer Mail List [email protected] http://www2.osgi.org/mailman/listinfo/osgi-dev
