[mod_python] onunload ie mod_python

On 28/10/2005, at 1:40 AM, Nick wrote:

> Graham Dumpleton wrote:
>> On 27/10/2005, at 6:31 AM, Nick wrote:
>>> See also  
>>> http://developer.apple.com/internet/webcontent/xmlhttpreq.html for  
>>> another solution (there are many resources for this information, but 
>>> I  like that one best).
>>>
>>> SimpleXMLRPCServer is a standard module in Python, and I don't think 
>>>  it's as complex as you might think, but the above solution might 
>>> yield  quicker results for you.
>> As far as integrating XML-RPC into mod_python, you are possibly better
>> of using the xmlrpclib.dumps() and xmlrpclib.loads() functions 
>> directly.
>>  From memory this gives better control over handling errors. At least
>> I know I had a good reason at the time which I can't think of right 
>> now.
>> Possibly because overriding SimpleXmlRpcServer didn't provide a good
>> way of returning a 404 when function didn't exist.
>
> I don't believe you must return a 404 when a function doesn't exist, 
> but only when a service doesn't exist.  From the spec:
>
> "Unless there's a lower-level error, always return 200 OK."
>
> A non-existant function would return a fault code.

One of the problems with the XML-RPC specification is that there is
no concept of an error origin. All error fault codes fall in the same
range. The SimpleXMLRPCServer if I recall properly returns a fault
code of 1 for missing methods or any internal problem it itself may
have. This is next to useless when you consider that any application
code usually starts error numbers at 1 as well, one just can't
distinguish errors very easily. It is thus helpful to at least use
a distinct error for not being able to map a callable function in
the first place.

I can't remember exactly, but there may have been other reasons as
well. It was all about the time when I found the security hole in the
SimpleXMLRPCServer implementation so it may have been a general
distrust of the code. :-)

Graham