[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [xmlblaster-devel] Re: [xmlblaster] Bugging me.



Martin Johnson wrote:

Hi Marcel



Ahh, it would be great if you could add your Windows C++ compiler settings
to our build.xml file



No problem, a week or so, it is turning into a very busy work week for me.



Possibly it would be nice to allow using our C++ code to use
lowlevel the SOCKET C-lib one day in future - this way
C++ clients would use xerces and STL and www.boost.org Threadlib
and other nice C++ libs but can choose lowlevel between CORBA or
SOCKET.



I can see the point in an embedded situation, I not even sure that that the full socket api is available on those types of platforms (select, ioctl etc) so dont limit yourself to a common denominator.

I think it would be good/wise to keep the underlying transport mechanism as
pluggable as possible. I think xmlBlaster is 'great' but the strongest
reason I considered xmlblaster is on the client side I already had an
established application and it is all on the strength of CORBA idl that I am
able to work with xmlBlaster seamlessly without adding additional untrusted
comms code. I think the same may stand for a socket based apps, what do you
think?

Yes CORBA is the best tested protocol with xmlBlaster
and it seems to be very stable and very high performing - i never had
any problems with it concerning reliable communication.

The xmlBlaster SOCKET protocol is very simple and allows
tunneling firewalls and with some tricks even proxies.
It is not very mature, but remaining bugs will quickly disappear
as it is very simple.
Also the performance needs to be tuned, as i expect it should
by possible to outperform CORBA with our SOCKET in future.

Have you heard of ACE http://www.cs.wustl.edu/~schmidt/ACE.html

Yes, great tools for C++.
But unfortunately it is a big beast and not very slim any more.
Binding xmlBlaster to ACE i would like to avoid as we force
all C++ users on ACE as well.
Our C++ client should be tiny and portable.
If users add ACE to our C++ code it should be there decision and
not enforced by us.


just my two Cents,

Marcel