Re: [xmlblaster] XmlBlaster error code discussion

[Marcel Ruff <ruff at swand.lake.de>]

> > > Hi folks,
> > >
> > > > after valuable feedback from Konrad Krafft, here is a new draft
> > > > of the future xmlBlaster error codes:
> > > >
> > > > http://www.xmlBlaster.org/xmlBlaster/doc/requirements/admin.errorcodes.html
> > >
> > >
> > > Hello guy, no comment, except like usually : a great and 
> > > comprehensive requirement.
> > > It just need to define each code ;o) 
> >
> > You are right, probably grep all existing XmlBlasterExceptions in our 
> > code 
>
> Is there a way to catch all of them ???
> Can I help ? 


Shure, it's dangerous to ask such a question ;-)
Making a list with all XmlBlasterException from our code as an
HTML table (or better requirements/xml table), and we could fill it step 
by step with errorCode definitions.
(do it only if you are keen on doing it ...)
Or any better idea?


> > and try to list them with a suitable errorCode could be an approach?!? 
>
> Perhaps there is some error code nomenclature in Corba ? in Http 1.1 ? 

(1) CORBA has 2 categories

 1. SystemException
 2. UserException

If we look a the SystemException they have constants defining the status:

 BAD_PARAM
 NO_MEMORY
 IMP_LIMIT
 MARSHAL
 NO_RESPONSE
 BAD_CONTEXT

and some more.

Additionally they have a CompletionStatus:

 _COMPLETED_YES
 _COMPLETED_NO
 _COMPLETED_MAYBE

but i don't know how this makes sense ( i did not search the internet
about this topic yet).
If an exception is thrown and we have a COMPLETED_YES
why would we throw an exception then??
A state COMPLETED_MAYBE, what does this help?
Can anybody shed some light on this??

Conclusion:
---------------
Corba has a similat approach like we do.
Konrad  keeps on telling me how nice the ORA errors of Oracle are
if you get  one you just surf the internet to find many answers.
Our more fine grained errorCodes (than CORBA) allow Konrad
to surf the internet when xmlBlaster has an error, like he does with 
Oracle...
-> bright future for him , eeehh

(2) HTTP style

HTTP has a nice erroCode based on numbers where 4xx numbers are
gouped logicall and 5xx number are dedicated to another error group.
This is similar to our approach.
But an "200 OK" makes in our context no sense (an Exception
throwing OK would be strange)




> Cyrille