15 May
2013
15 May
'13
6:04 a.m.
Dear Yusuke,
thanks for the advice to your paper ... great work!
We experimented also with EXIP on the Zolertia Z1/Telos B, but the memory usage is not
acceptable for these hardware platforms as you already mentioned in your paper, too.
You have found a good way to shrink the memory usage of EXI, but its still too
"heavy" to run aside of uXMPP on these mentioned hardware platforms.
So, we decided at the moment to avoid the use of EXI, but if there is a library which
really requires less than 10kB on the msp430-based hardware platforms we should definitely
consider the use of EXI in uXMPP !!!
I'm really looking forward to these days ;)
Best regards,
Ronny
On Wed, 15 May 2013 16:07:42 +0900
Yusuke DOI <yusuke.doi(a)toshiba.co.jp> wrote:
Dear Ronny,
Interesting work, thanks for publications.
(2013-05-15 02:07), Ronny Klauck wrote:
--
Dipl.-Inf. Ronny Klauck
Brandenburg University of Technology Cottbus
Computer Networks Group
http://www-rnks.informatik.tu-cottbus.de/~rklauck
But we did a lot of (unfortunately unpublished)
performance tests
with other EXI implementations on the Zolertia Z1 sensor node, which
showed that the compression of EXI for XMPP messages is high but the
implementation effort and the memory consumption as well. So we
decided to optimize XMPP messages and there behavior in a smart
object network directly to have more available memory for XEPs! This
solution and the implemeted XEPs can be found as well in the
mentioned papers/journal.
Have you tried EXIP? As I checked out ws4d-uexi, I found there are imported sources from
EXIficient (JAVA) -- maybe not a best codebase for embedded use.
In theory (and some of our practices), adequate design of schema and protocols may reduce
required memory for EXI down to less than 10 kBs.
If you have interest, some of ideas may be found in the following paper.
http://ieeexplore.ieee.org/xpl/articleDetails.jsp?arnumber=6402307
Yusuke