Re: [Netconf] NETCONF over SOAP issue withdraft-iijima-netconf-soap-implementation-10

["Bert Wijnen \(IETF\)" <[e-mail]>]

So it seems that neither the WG chairs, nor the Netconf-over-soap 
authors/editors.
nor any other WG member had noticed this. Strange. I will hacve to go and 
re-read.

Personally I am somewhat worried if we publish "implementation experience of
Netconf over SOAP" if it is not interoperable with the NETCONF over SOAP
standards track specification!!!! Even if that is spelled out clearly.

Does anyone else in teh WG have an opinion (or care)?

Bert

----- Original Message ----- 
From: "Romascanu, Dan (Dan)" <[e-mail]>
To: <[e-mail]>
Cc: <[e-mail]>; <[e-mail]>
Sent: Monday, September 01, 2008 4:37 PM
Subject: [Netconf] NETCONF over SOAP issue 
withdraft-iijima-netconf-soap-implementation-10


> During the IESG review of draft-iijima-netconf-soap-implementation-10
> (AD-sponsored independent submission targeting Informational RFC) Pasi
> Eronen raised the issue of the usage by the implementation described in
> the document of the cookie field inside the HTTP header to map incoming
> requests to NETCONF sessions.  This makes such an implementation
> implementation to be an alternative SOAP binding for NETCONF which does
> not interoperate with RFC4743 compliant implementations.
>
> In the IESG discussions we chose to approve the document as
> Informational but make clear in the text that such implementations are
> not interoperable. We would like however to have the NETCONF WG look at
> the proposed text and make sure that we got it right and especially that
> RFC4743 is not in error.
>
> If you believe that there are any problems with the proposed text please
> let us know until 9/10 the latest.
>
> Proposed Notes to the RFC Editor:
>
>
>
> 1. In the Abstract Section:
>
> OLD:
>
>  This document describes how the authors developed a SOAP (Simple
>  Object Access Protocol)-based NETCONF client and server.  When SOAP
>  is used as a transport protocol for NETCONF, various kinds of
>  development tools are available.  By making full use of these tools,
>  developers can significantly reduce their workload.  The authors
>  developed an NMS (Network Management System) and network equipment
>  that can deal with NETCONF messages sent over SOAP.  This document
>  aims to provide NETCONF development guidelines gained from the
>  experience of implementing a SOAP-based NETCONF client and server.
>
> NEW:
>
>  This document describes how the authors developed a SOAP (Simple
>  Object Access Protocol)-based NETCONF client and server. It
>  describes an alternative SOAP binding for NETCONF which does not
>  interoperate with an RFC4743 conformant implementation making
>  use of cookies on top of the persistent transport connections of HTTP.
>
>  When SOAP is used as a transport protocol for NETCONF, various kinds
> of
>  development tools are available.  By making full use of these tools,
>  developers can significantly reduce their workload.  The authors
>  developed an NMS (Network Management System) and network equipment
>  that can deal with NETCONF messages sent over SOAP.  This document
>  aims to provide NETCONF development guidelines gained from the
>  experience of implementing a SOAP-based NETCONF client and server.
>
> 2. Add to Section 1.2 the following paragraph:
>
>  This  document describes an alternative SOAP binding for NETCONF which
>  does not interoperate with an RFC4743 conformant implementation as it
>  relies on cookies used on top of the persistent transport connections
>  of HTTP. This is provided for information purposes only based on the
>  implementation experience of the authors.
>
> 3. In Section 3.1.2
>
> OLD:
>
>  In [RFC4743], HTTP is specified as an option of an underlying
>  protocol for NETCONF over SOAP.  When HTTP is used for that purpose,
>  it is also specified that a NETCONF session will be supported by an
>  HTTP connection.  However HTTP is a stateless protocol; that is, HTTP
>  cannot process a user's request according to the state resulting from
>  the user's previous request.  Unless the state is kept at the HTTP-
>  level, a different NETCONF service provider will be invoked every
>  time the NETCONF application sends a NETCONF message to the NETCONF
>  service provider.  To ensure that the same NETCONF service provider
>  is used every time the NETCONF application sends a NETCONF message,
>  the state of the HTTP connection must be maintained.  Accordingly, a
>  cookie field inside an HTTP header was devised for maintaining the
>  state of an HTTP connection.  We therefore used such a cookie field
>  to maintain the state of the HTTP connection over which the NETCONF-
>  session maintenance is ensured.
>
> NEW:
>
>  In [RFC4743], HTTP is specified as an option of an underlying
>  protocol for NETCONF over SOAP.  When HTTP is used for that
>  purpose, it is also specified that a NETCONF session state is tied
>  to the state of the underlying transport (TCP) connection (just
>  like in NETCONF over SSH [RFC4742] and NETCONF over BEEP
>  [RFC4744]). However, HTTP itself is a stateless protocol, and many
>  server implementations process user requests independently of
>  previous requests received over the same transport connection.  To
>  simplify implementation of the NETCONF service provider, we used
>  the cookie field inside the HTTP header to map incoming requests to
>  NETCONF sessions. Note that this means our implementation actually
>  uses an alternative SOAP binding for NETCONF which does not
>  interoperate with RFC4743 compliant implementations.
>
> Thanks and Regards,
>
> Dan
>
> _______________________________________________
> Netconf mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/netconf
>
> 


_______________________________________________
Netconf mailing list
[e-mail]
https://www.ietf.org/mailman/listinfo/netconf