axis-c-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Aleksander Slominski <as...@cs.indiana.edu>
Subject Re: RFE: axis c++ engine tracing
Date Thu, 17 Jun 2004 17:19:25 GMT
John Hawkins wrote:

>
>
>Sounds a good idea. However, this must be internationalised.
>
>There are two considerations here:
>
>Debug Tracing - this is purely for developers to understand what went
>wrong. This is not something that should done lightly - if we do this then
>we should do it properly i.e. put in entry/exit trace everywhere or nowhere
>- there is nothing worse than dong a half job of trace ! Because this is
>for developers only there is no need to internationalise this.
>  
>
that is what i wanted: trace dispatching and execution of SOAP message.

>Informational "messages". I have to say that I think we would have very
>little, if any,  of these? 
>
agreed

>Most/all of our messages are, hopefully, now
>exceptions. In which case, at the client-side,  it would be up to the
>application writer to catch the exceptions and translate them to their
>language as suitable. The chances are they would want to alter/wrap the
>exceptions into their own format anyway. I think we should avoid putting in
>any informational messages at the client, if at all possible. Perhaps on
>the server side there is need for a "starting Axis server", "Stopping Axis
>server" message 
>
yes.

>There are plenty of opensource pkgs out there that do internationalisation
>and logging. I would point you towards this one  -
>http://oss.software.ibm.com/icu/ - which we've looked at before (oh, it
>happens to come from IBM too ;-)
>  
>
:) isnt it very complicated?

>Now, going onto another favuorite topic of mine Aspect orientation. It
>would be possible to make entry and exit trace an aspect and apply it at
>build time ! This would mean, if we needed to change the trace (for
>whatever reason) in the future we would just change the aspect. It also
>means that we can keep the codebase easier to work with and avoid people
>having to swim through the trace to find the functional code.  I don't know
>how good aspect C is now adays but I'm sure it would be up to this simple
>job (aspect people always cite entry and exit trace as their samples!)
>
>  
>
i am not that big on tracing entry/exit but more interested to add some 
tracing/logging of key data/control points such as decision to what 
service message is dispatched (or dropped). i spent hours debugging this 
in AXIS-C++ and if we could save time for future developers working with 
AXIS-C++ ...

thanks,

alek

-- 
The best way to predict the future is to invent it - Alan Kay


Mime
View raw message