Thanks Carl for pointing me to the Google group and the wiki!

The approach to compile and run the existing BOINC client with the NDK
definetly has something to it. As I understand it spares us porting the
client to Java and should also perform nicer.

Within this architecture the part of the implementation on top of the Java
- SDK would
a) own the GUI
b) enforce restrictions (e.g. only accept jobs when on wifi, calculate only
when charging, ...)
and c) operate the communication

Since native ARM code runs in the same instance of DalvikVM as the GUI the
TCP connection could (and should?) be replaced by JNI.

Is it fair to say, that the communication between BOINC servers and the
client has to go through Java, too?

I haven't worked with the NDK yet, but - with Peter's work on BOINClite in
mind - would it might make sense to start from scratch with a native code
implementation of the BOINC client that fits into Android's security model?

-Joachim

On Wed, Mar 21, 2012 at 8:54 PM, Carl Christensen <[email protected]> wrote:

> there's a google boinc-droid group:
>
> http://groups.google.com/group/boinc-droid
>
>
> also some new work / news here -- I know this is an important direction
> for BOINC to get on Android:
>
> http://boinc.berkeley.edu/trac/wiki/AndroidBoinc
>
> the efforts so far seem to be split into two "camps" -- make a native Java
> Android version of BOINC (BoincOid, AndroBOINC etc) --- use the NDK to
> compile the C++ BOINC & associated libs (me & Peter Hanappe in this
> direction) -- the latter would still require a Java/Android GUI to
> communicate, with a different means than the usual BOINC gui_rpc calls
> since (at the time I was playing with it) android didn't support the sysv
> network/memory stuff required.
>
>
> It's still up-in-the-air whether "traditional BOINC" (i.e. downloading
> apps and then running) will fit into the Dalvik VM esp the security model
>  (I only got it to work with a "rooted" phone & Android VM).
>
> we are probably at the point we should have a "
> [email protected]" list to get serious work done (there's
> just a private email list of interested parties right now) ?
>
_______________________________________________
boinc_dev mailing list
[email protected]
http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.

Reply via email to