Current Meeting Report
2.4.11 XML Configuration(xmlconf) Bof
Current Meeting Report
XML Configuration BOF [xmlconf]
Monday, July 15th, 1300-1500
Margaret Wasserman <firstname.lastname@example.org>
There is significant interest in the use of XML-based technologies to manage and configure IP-based networks and networking equipment. This BOF will discuss the high-level requirements for a network configuration mechanism, and will explore how the existing proprietary and standards-based XML technologies, such SOAP, WBEM and SyncML, could be used to meet those requirements. Based on that discussion, we will determine whether we should pursue any IETF work in this area.
- Introduction and Agenda Bashing -- Margaret (10 min)
- Requirements Discussion
- Operator Needs
-- Weijing Chen (10 min)
-- Ron Bonica (10 min)
- Use of Metadata in XML Configuration -- Ray Atarashi (10 min)
- Requirements Strawman -- Margaret (15 min)
- Survey of Existing Technologies -- Margaret (15 min)
- WBEM & CIM/XML
- Conversion of MIB Data Model to XML -- Andy Bierman (15 min)
- Discuss Further IETF Work (35 min)
Concepts and Requirements for XML Network Configuration
http://www.ietf.org/internet-dafts/draft-wasserman-xmlcomf-req-00.txt Towards XML Based Management and Configuration
http://www.ietf.org/internet-dafts/draft-goddard-xmlconf-survey.txt Guidelines for the Use of XML within IETF Protocols
Network Management Observations
http://www.ietf.org/internet-dafts/draft-bierman-nm-observations-00.txt Other Useful Pointers:
WBEM & CIM/XML
Intro & Agenda Bashing:
No agenda changes.
Weijing Chen presented overview of his needs (see slides).
Summary: A common XML schema would allow disparate backend systems to interface the networking equipment in a uniform fashion.
Ron Bonica discussed his use of Juniper's XML configuration facility.
Summary: The XML facility in Juniper is accessed through a CLI front-end. Both the CLI server and the XML server run on the box and communicate with each other. Also possible to access the XML server directly.
For more information about Ron's work with Juniper products, see the notes from the Ops Area open meeting.
Discussion: Questions raised about the assumption that XML would be used for access to SNMP data. Does SNMP really have access to all the data that is needed, etc. A: SNMP is being used for these things now, and XML would provide a more consistent way to access it.
Use of metadata in XML configuration:
Summary: Ray Atarashi presented her research into metadata and its applicability to configuration systems (see slides).
Summary: Margaret Wasserman presented the requirements strawman with an emphasis on open issues and questions (see slides).
Question: Change notification is out of scope? A: no, it is in the requirements. However, as few mwchanisms as possible need to be built at once. Open for discussion
Comment: why use XML? Just to have XML inside? A: XML as a data representation method has advantage to be human readable and machine parsable
Comment: need to define first what is the problem to be solved
Rob Austein: At the IAB NM workshop requirements were enumerated. On Thursday there will be a presentation on this in the IAB Plenary
Dave Harrington: The SNMP framework document explains the components of a management solution. This model can be used in order to decide what XML can be used for A: Not the goal to reinvent SNMP in XML.
Dave Perkins: Two ways of presenting the info - DDT and schema
Collin Jenings: Why would the solutions isomorphic with SNMP? There are some other requirements that need to be addresses such as revision control and change synchronization. Margaret: Can you send details to the list?
Randy Bush: Agrees. Need for transaction model. Where does XML give leverage over these items?
Internationalized strings - need volunteers to contribute to requirements in this area.
Security - need to understand what requirements actually are.
Steve Bellovin: authorization and operational model needs to be defined. What needs to be secured? Do we need object security, or just transport security?
Randy Preshun: on previous point of transaction modeling - draft-preshun-webdav... talks about a number of these issues plus about security model.
Eliot Lear: Many vendors do some form of XML configuration, they are doing it differently, need a minimal level of interoperability M: would like to start with the transport, but other people would like to define a schema first
Eric Rescorla: Why would it have to be XML data? Any data could be sent over a secure channel. M: might be useful to put some formatting so that a client do some diff-ing and know what pieces of info to deal with. XML-izing the SNMP might be in the minds of some people
Peter Lothberg: Runs a "small" network. Wants to talk to network elements as a human being; also wants machine language to talk to routers from management system; using XML seems to do the job; make it secure
Harald Alvestrand: There are specifications to translate between languages. This is not the issue. Focus on making the language useful for configuration. Easy things should be easy, hard things should be possible.
Dave Durham: Using MIBs is polluting the data model because of the lack of a transactional model
Sharon Chisholm: SNMP will still be used for monitoring thus need to do some mapping between XML and SNMP
Will an XML data model be easier to instrument? To be worth-while and gain acceptance, needs to be extensible and cheap to implement
Shai Herzog - nice to monitor things that we configure, configuration needs to say if transaction is done, this is not monitoring by transactional configuration
Comment: disagrees that configuration can be done the same for short lived configuration and provisioning
Other data representation? - would be subject of a separate BOF
Randy Bush: Should be discussing schema,not transport!
[Discussion was cut-off to allow presentations to continue]
Skipped over survey presentation, due to time pressure (see slides if interested).
Conversion of MIB Data Model to XML:
Andy Bierman presented ideas about this conversion (see slides).
Discussion continued on whether to define a transport, define a schema or do both. Further discussion of how/if XML helps to make any of this easier or more useful for operators.
No consensus was reached during the meeting about how/if work in this area should continue in the IETF.
Ran out of time, with discussion to be continued in the OPS Area open meeting.
XMLConf Follow Up
Use of Metadata in XML Configuration
XML Network Management Protocol and Interface
- Ray S. Atarashi
- Yoshifumi Atarashi