axis-java-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Pridemore, Russell (MAN-Corporate)" <>
Subject RE: Abort a axis handler chain
Date Wed, 12 May 2004 20:50:34 GMT
It sounds like what you want is HTTP DIGEST authentication, which supplies
a client certificate outside the realm of SOAP.  Or am I lost (again)?
Have you looked into it?  Doing so would not simply drop the connection,
it would return a standard HTTP 401 (unauthorized), however...


-----Original Message-----
From: Robert Mecklenburg []
Sent: Wednesday, May 12, 2004 4:43 PM
Subject: RE: Abort a axis handler chain

RM> I want to implement a service that authenticates the caller (using a
RM> certificate in the message).  If the user is not authorized to invoke
RM> the service, I want *no response* sent at all.  No return value, no
RM> exception, nothing.  Is this possible?

Anne Thomas Manes writes:
ATM> If the service is defined as a request/response pattern, then I don't
ATM> that's possible. You should return a SOAP Fault, but you might consider
ATM> sending back an HTTP error instead.

Nelson Minar writes:
NM> I agree with Anne, this is a somewhat unnatural thing to want to do.
NM> The simple thing would be to simply return a SOAP Fault: the caller
NM> would get a 500 error back with some basic data in the body.

We are very concerned about security in our server.  If a "bad guy"
attacks our server we don't want: a) to reveal any information about
the failed authentication, or b) to place any further burden on the
underlying network.  The response could contribute to a DoS attack,
for instance.

I'll investigate your suggestions (and Robert Lowe's), thank you.  In
the meantime, we came up with these two possibilities:

1. We could spawn a thread to process the soap message and
   pre-emptively terminate the thread in the event of an
   authentication failure.

2. Re-write the return address in the soap header to route the
   message into the bit bucket or to another soap client of our own
   choosing (which would dump the response in the bit bucket).

Do either of these sound promising?


View raw message