Return-Path: Delivered-To: apmail-xml-axis-dev-archive@xml.apache.org Received: (qmail 99953 invoked by uid 500); 13 Mar 2001 00:55:44 -0000 Mailing-List: contact axis-dev-help@xml.apache.org; run by ezmlm Precedence: bulk Reply-To: axis-dev@xml.apache.org list-help: list-unsubscribe: list-post: Delivered-To: mailing list axis-dev@xml.apache.org Received: (qmail 99929 invoked from network); 13 Mar 2001 00:55:42 -0000 Received: from georgia.yamato.ibm.com (203.141.89.181) by h31.sny.collab.net with SMTP; 13 Mar 2001 00:55:42 -0000 Received: from e022f2n10.jp.ibm.com (d22relay02.yamato.ibm.com [9.68.14.53]) by georgia.yamato.ibm.com (8.11.1/3.7W/NG3.6) with ESMTP id f2D0tMG18728 for ; Tue, 13 Mar 2001 09:55:22 +0900 Received: from e022f2n7 (d22hubm7 [9.68.14.51]) by e022f2n10.jp.ibm.com (8.8.8m3/NCO v4.95) with ESMTP id JAA20064 for ; Tue, 13 Mar 2001 09:55:22 +0900 Importance: Normal Subject: Re: Pass-Through Axis Nodes To: axis-dev@xml.apache.org X-Mailer: Lotus Notes Release 5.0.3 14 April 2000 Message-ID: From: "Yuhichi Nakamura" Date: Tue, 13 Mar 2001 09:55:20 +0900 X-MIMETrack: Serialize by Router on D22HUBM7/22/H/IBM(Release 5.0.3 (Intl)|21 March 2000) at 2001/03/13 09:55:21 AM MIME-Version: 1.0 Content-type: text/plain; charset=us-ascii X-Spam-Rating: h31.sny.collab.net 1.6.2 0/1000/N Hi Pawel, Your problem can be solved in the current Axis architecture. Some Axis nodes can be intermediaries. BTW, do we have intermediary samples in the cvs tree? Comments in line: >Gentlemen, >I am currently exploring the ways to protect my web services communications >from unauthorized access, spying eyes and bulkiness of the messages. There are >few solutions already on the net - like Digital Signature extension from IBM or >simple authentication in Axis. They seem however to be very basic considering >the potential requirements for the data being sent. Digital Signature and basic authorization (SSL-based) are a good starting point. I guess. >The chaining concept presented in the Axis architecture document would be >perfect to employ certificate based authentication, non-repudiation, etc. There >is a problem however with the client side where the chain elements to e.g. >encrypt message are missing. The digital signatures demo from IBM solves the >problem using a special SOAP service to wrap the original message but that does >not provide an elegant chaining solution. The nice solution would be to have, >except Axis end-nodes, Axis pass-through nodes. Those nodes would serve for >example as gateways for communication out of the company. The additional bonus >would be an ability for strong enveloping - user would use SSL to send the >request to the nearest pass-through node. Leaving the node the SOAP packet may >be wrapped multiple times in the intermediate nodes for other purposes, for >example separate encrypted and signed POs may be bundled in one message. On the >receiving side a similar node would route the requests or their parts to the >respective departments. We should have digital signature handler, basic authentication handler, encryption handler, etc. They can be placed in ANY Axis nodes, including intermediaries. >I have tried to use existing framework to develop the pass-through nodes but >there is a problem with the early deserialization. The wrapping services >usually know nothing about the wrapped content except it is xml compliant. >That >would require either keeping the content in xml or using fake >serializers/deserializers to wrap the content into java classes. While the >latter is possible to do it may however interfere with deployment tools in the >future. >As of the question in the doc if the messages should serialize/deserialize >themselves or use external classes I think the external classes are a better >solution. For now I am able to create services with quite complex interfaces >using the Castor generated classes (http://castor.exolab.org/) . After creating >the xml schema for the message it takes literally minutes to have the shell for >the services with all the serializing/deserializing classes generated. It would >be however difficult to integrate Castor with Apache SOAP should the type info >mapping require separate serializer/deserializer for each of the types mapped. You mean "current Axis." If you report your problem in more details, some of Axis members will take a look at it. I feel that serialization/deserialization is independent of security issues. If you addressed "message" instead of "rpc", would you still have the security issue? >Do you think the existing Axis architecture may be extended to accommodate >pass-through processing as well ? Yes. >Best regards, >Pawel. > Yuhichi Nakamura IBM Research, Tokyo Research Laboratory Tel: +81-46-215-4668 FAX: +81-46-273-7428