Draft Minutes

[Steve Kille <[e-mail]>]

Please send comments by tomorrow.

Steve

Minutes of the OSI-DS Working Group, Columbus, Ohio.
Minutes taken by Ed Reed <[e-mail]>

29 March 1993 Started at 13:40

Introduction
        - Agenda
        - Minutes of Washington DC Meeting
            Accepted without comments
        - Other matters

Liaisons
        - WG-NAP (Erik Huizer)
            There are 3 task forces working:
            Work that Panos-Gavriil Tsigaridas is doing.
            Data Management - how to get data into the directory, and
                keep it up to date and accurate - a first draft is avail.
            Legal and Privacy issues - going to publish first results
                as an analysis of Dutch and other regulations recently
                published.
        - NADF (Tim Howes)
            Decided to open to users, but as non-voting members.
            Piloting continues.
        - DISI (Tim Howes)
            Reformed under Integrated Directory Services WG, with the
            addition of Whois++, want to address general directory
            issues.
        - AARNet (Mark Prior)
            One of the divisions of Telecom have joined the pilot.
            Currently doing a Whois++ startup.
            Trying to get a new binary distribution of ISODE 8.0 ready.
        - Paradise
            The First project ended with 1992
            Next Project (transition) will run 16 months
            Now also has Inrea as a partner.
            Using Quipu and Pizzaro implementations.
            DE now will perform very broad (c=??/cn=name) searches.
            China, Slovenia (means ISODE 8.0 required), Croatia, Slovakia
                have joined.
            A question was raised about the issue of intellectual property
                rights associated with DE and other Paradise tools.
        - NREN-NIS (Sri Sataluri/Mark Kosters)
            Internic will begin providing services April 1. - Internic
            personal listings will be available.  
            Will provide DE access, and other X.500 services and WAIS.
            Plan to make WAIS info avail via X.500, too, but not immediately.
        - DOD
            Defense messaging system will take over the old switches
            including Autodin.  Will be X.400 based, and are building an
            X.500 directory support.
        - Integrated Services Panel (US/GSA) (Thomas ...
            There's a newsletter describing efforts available.
            Directory services now are flat file, with X.500 direction.

Progression of Standards (Erik Huizer)
    Published the Strategy doc.
        There was a long discussion by John Curran but he was unable
        to be here to discuss.
    LDAP was published - or maybe it hasn't, but it's left the IESG.
        There may be a block of some sort - Tim Howes took an action to
        see if it's being held up in the IAC.
    No other documents are pending immediate progression.
    DSA Metrics will be discussed later.

Experiment Progress
    DIT Counting
    Character Set
    JPEG
        Concluded at previous meeting, other than for publication of
	new approach in updated RFC 1274
    QOS
        DE QOS:  - Paul Barker
            Feature Added to latest DE
            Those familiar with Directory Knowlaged
                * Which org's data likely to be available
                * Likely to be returned tolerably quickly
                * attempts to provide naive user with some info
            Doesn't follow OSI-DS 15
                * coverage - 1 out of 62 GB orgs have QOS attributes
                * Data vs DSA - emphasis should be on data avail,
                    rather than DSA avail
                * Response time - no attempt made in OSI-DS 15 to indicate
                    likely response time
                * Credibility - values self-assigned.

            Approach used
                DE uses simple database of information availability and 
                response times
                    - "result" of each query added to QOS database
                    ((query times > threshhold time) &&
                        info for that query is database))
                    users told how long query usually takes
                                    or
                             <information lost that I didn't get copied>

                Shortcomings of current approach
                    Database built only from 'simple' query mod
                        power searching provides much more information
                    No account taken of when a query is made (time of day)
                    More hysteresis is needed
                    Abandons are not recorded
                    no timestamping of information
                    Tuning tools are needed
                    Database is too simple at the moment.

                Have not concluded that the draft OSI-DS 15 should be abandoned,
                if more of the values specified there are in fact implemented.
                Question is whether DSA and DUA implementors will build OSI-DS 15
                approaches, or not.

                There is some overlap between this experiment and the MADMAN 
                efforts.

                Gavriil Tsigaridas reported some of their efforts have
		raised an
                issue with DS-15's use, or lack of, object type data
	        in the QOS
                database.  One approach is to just record information
		about times to find people.

                Information is probably only of interest at the local level - 
                views are too different from other places via other access 
                mechanisms.

                There is some value to share the implementation approach taken,
                but this is an experiment, still.  


            Seems like we've gone far enough on DS-15, and we should look
            in some other direction.  DS-15 is complementary, but should
            be extended with interface native information which doesn't
            belong in the directory.

            Paul will see if he can make what he's done available for others
            to implement.

            A poll of the group indicated continued interest in developing
            OSI-DS 15, but no volunteers to be the editor.

Schema WG
    A previous meeting chartered a small subgroup to look at this.

    That group never got together.  There have been other issues (JPEG, etc)
    which have come up needing help, too, but still no volunteers to edit.

    Looking for a volunteer.

    No Volunteers.

Panos Gavriil Tsigaridas
    Asked people to please read his document.
    Applications need the ability to use a common respository for information
    about management information, there'd be a valuable synergy.

Charter Review
    Draft Charter Steve published
    Another that Erik published.

    Steve doesn't think it makes sense to put things into the charter
    which need to be done, if there's not support from the WG members to
    do them.

    Steve proposes 4
        - Liaisons
        - Schema Coordination
        - DSA/DUA Metrics
        - IP address representation
    
    Erik points out we need to be stricter in our procedures and resources
    as the IETF grows - specifically with regard to charter and time
    schedules.  Only if we have concrete objectives and times will our
    charter be renewed.

    Erik's list includes
        - Non-white pages use of the directory
        - test strategies
        - schema management 
        - guidelines for technical implementation, migration to 1993, 
            and database coupling

    Paradise has a objective including interoperation of directory services.

    Interoperation is more properly a target of pilot projects, with which
    we want to liaison, but that should not be part of this groups charter.

    Perhaps if we defined where the holes in the standard exist which
    preclude interoperability, and publish RFCs to fill the holes, then
    at least we'd have a unified face to the implementors.

    For instance, Siemans has delivered RFC based product, which goes
    beyond the OSI Standard, when pressed by pilot managers.

    To some extent, this seems to be a necessary activity, in spite of the
    continuing claims by vendors that the extentions are non-standard, and
    will be obsoleted by the next standard.

    (Erik) - IDS will focus on general problems relating to directory
         services, while OSI-DS will focus on X.500 specific issues.


Metrics
    Roland - Have been testing the Siemans DSA.  Has also had to look
    at interopability testing.

    There are holes in the standards - schema handling, access control, etc.

    There appears to be holes in the metrics, to - they report good results,
    when you know there are problems.

Paradise - Paul Barker
    Discussed new tools in more detail - which will be available shortly.

    Archie-like service based on X.500...begins with a leap of faith that
    it makes sense to record information about documents in the directory.
    The presentation provided an overview of the approach to be taken.

    Sense of the group was that the WG should take the project being done
    as a work item.  Paul to edit the papers he's done towards an RFC.

Representing WHOIS data in the X.500 Directory (Sri Sataluri)
    Objective is to provide access to information about network entities
    and to define a schema for representing that data.

    Concern expressed that approach may simply be replicating a centralized
    database, and not really distributing it - but there was disagreement
    with that concern.  

Representing IP information in the DIT (OSI-DS 32-39) 
    - Glenn Mannsfield and Thomas Johannsen

    Presentation included background, problem discussion and description
    of a proposed solution.

    Objective is to provide a distributed map of the network.

    Not only topology, but the policies, costs, services, properties,
    administration and management attributes, and contacts.  Many kinds
    of applications can use the information, but network management is
    the main thrust of the effort.

    CONMAN Project is addressing configuration management.
    SOFTPAGES Project is addressing cost computation, using the configuration
    information from CONMAN, etc.  In addition, file server contents is
    indexed in the directory.

    Sense of the group was that the WG should be dealing with the
    problems described here.



		Abstract:Charting Networks in the Directory.
                         work in progress at AIC, WIDE, Tohoku Univ.

There is a dearth of information about the network
	- Interconnections, policy of transit n/w's, contact persons, ..
The present status of the n/w info is piecemeal and diverse
	- geographical separations [ the various NICS, ...]
	- specific Usage oriented  [ DNS, whois, ....     ]
A Unified view is proposed- something like a global annotated n/w map
	showing interconnections and their properties and policies
	        the functions/services of the elements
	        admin/mgmt related info
	form the base of Directory Services
		name , address , manager, policy, route, ...
	
The Map may be used for 
	Conf mgmt : see n/w configuration, designing/administration/planning
	Route mgmt: checking optimality of paths, support route servers, ...
	Fault mgmt: alternate paths, ..
	Service mgmt: information on servers/services, Managers, users, 

By definition the Map is Huge, quasi-static, geographically distributed and
        requires distributed control & maintenance

X.500 based distributed directory provides the base for such a map

Points Addressed in the Proposal:
	-Scalability, distribution of control & maintenance, preservation of
	 admin/political boundaries < X.500 based model
	-Simple representation      < should be close to the real world
	-Minimize data duplication  < images like organizationRole to be used
	-Use existing services/info [ DNS, NIC ] for bootstrapping
	-Address evolving technologies/problems [ supernetting, ..]
The network Map:
	- comprises of networks, nodes, interfaces
Images:
	- allow several functional images of the same physical n/w
	  OSI/IP/SNA descriptions of the same n/w is possible

The Appplications that are coming up:
	- ConMan Project
		- Configuration info supplements other mgmt info
		- Displays map, finds manager who should be contacted 
	        - Suggests bypasses in case of problems 
        - SoftPages Project
		- Target is to optimize document retrieval
		- The "Map" gives the cost [function of speed, tariff, ...]
		  from the user to the ftp servers
		- The "Map" also contains info about the servers and contents
		- The "cheapest" server from the user is found
	- NIC info server
		- provides a single-window whois-type service

Status:
	- Pilots have been implemented         [ Thomas will present ]
	- experimentation has been carried out [ Thomas will present ]

Plan of Action
	- develop strategies/tools for populating the Directory
	- take the pilot to wider [ national -> international] arena
		link NICs, Maps
	- develop nice UAs, applications
Time Frame
	- Next IETF:
		More Results, population, coverage, usage. 
		Bootstrapping strategies.
In the minutes I sent you, I included some content from presentations - Here
is input from Thomas on his presentation - could be included as is, or
as an addenda.


Notes to talk of Thomas Johannsen:

Implementation of OSI-DS 37-39 in national pilot done.

Aim: representing
 - networks
 - IP numbers
 - ftp archives

Interaction of following information supported by references 
and tree structure:
 - white spaces namespace
 - network descriptions
 - IP number namespace
 - DNS namespace

1) Networks in the Directory: 
   Populating

   => in general no config database available
   => upload existing data from WHOIS, BITNET.NODES, DNS...
   => gathering data "by hand" from network admins, ...

   ==> Technical and administrative support needed.

   NIC support

   experimental upload of parts of JNIC-WHOIS database into
   X.500 done. 

   Autoconvert for part of JNICs database (IP numbers 133.*.0.0) produces:

   X.500 object  	number 

   organization  	194
   organizationalUnit 	247
   pilotPerson 		429
   IPnetwork  		228
   IPgroup 		228

   => uploading WHOIS to X.500 helps populating white pages
   space, too!

   Problems of autoconvert:

   - non-unique use of org-names
   - addition of organizational entries over DSA boundaries difficult
   - X.500 access rights

   X.500 based whois responder as user agent provides access to
   white pages and non-white pages information. Send mail to
   [e-mail] with subject 'help'.

 2) Pilot Directory configuration (for OSI-DS-38/39)

   3 DSAs form isolated pilot world for experiments
  
   number of objects in pilot DSAs:

   object 		Sazae 	Saki 	Guppy 	total
   organization 	8 	3 	193 	204
   org.-Unit 		19 	11 	244 	274
   pilotPerson 		34 	7 	416 	457
   network 		13 	0	0 	13
   ipNw'Image 		41 	8 	202 	251
   node 		71 	0	0 	71
   ipNd'Image 		80 	5 	2 	87
   port			49 	0 	0 	49
   ipPortImage 		83 	8     	3 	94
   del.Block		0	0	24 	24
   ipGroup		0	0	244 	244
   ipReference		0	0	65 	65
   fileServer 		8 	4     	3 	15
   file 		12287 	2839	7895 	23021
   total		12693 	2885 	9291 	24869

3) Soft Pages Project (OSI-DS-39)

   aim: reduce unnecessary ftp traffic

   - provide help for efficient and network optimized file retrieval
   - use of network connection properties
   - representation of contents of fileservers in the Directory

   Maintaining SoftPages:
   - initial loading of all filenames into the Directory (once
   only)
   - daily addition and deleteion of new or outdated files 
     (number of files depends on size and activity of fileserver)
   - final deletion of all filenames from Directory when fileserver goes
     out of operation (once only)

   => can be done by crontab job which reads ls-lR, builds diff
      to previous entry and has DUA functionality (add/del in DIB)

   estimated size of ftp servers to track: typically 3000 -
   10000 files; with several tens to hundreds changes per day.

   Bulk loading tests performed on pilot DSAs. 
   Summary: addition of 10000 file objects takes approx. 30 minutes
            daily updates takes about 1 minute	    

   Using SPP

     Searching filenames as
      - full match,
      - leading substring match,
      - non-leading substring match.

     Result of tests: searching one object takes about 1 second 
     for amount of up to 10000 objects in one ftp-mirror.

   Investigating impact of SPP


   cost = f (speed, traffic, charge, priority)

   cost calculation in experiment done by using ftpd-logs and
   simplified backbone network map of Japan Internet.

 	- randomly choosen logs of 5 ftp sites
 	- scanned about 35000 anonymous get operations
 	- checked for filename, size, date against 45 ftp sites
 	- IF file was found in ls-lR of a cheaper ftp site THEN
          non-optimal retrieval.


   results expressed as non-optimality ratios:

    no. of files  32 % 
    bytes         38 % 
    total cost    50 % 

4) Summary

Non-white pages X.500 usage by several applications, 
e.g. NIC control, ConMan project, SoftPages project.


---------------------


    Subgroup of volunteers agreed to meet over dinner and plan work.  
    Members include:
        Paul Barker
        Tim Howes
        Thomas Johannsen
        Mark Knopper (silent volunteer) (missed dinner)
        Mark Kosters
        Ruth Lang
        Sylvain Langlois
        Bruce Mackey
        Glen Mannsfield
        Ed Reed
        Sherry Repucci
        Sri Saltaluri
        Mark Smith
        Scott Williamson

    This group (less Mark Knopper) had dinner, identified a list of documents
    to be published, and accepted volunteers to edit the docs.

    The list of documents identified, and the people agreeing to work on them:

    Roadmap (Steve H-K)
    IP Addressing Schema (Glenn, Thomas, Mark Ko, Sri)
    Network Objects Schema (Thomas, Sri, Ed, Mark Ko.)
    RFC1279 Revision (Mark Ko.)
    Naming Layout (Sri)
    Transition Plan for Existing Services and Deployment (Scott, Glenn)
    Business Process Model (Operations Guidelines) - Glenn
    Security and Privacy (Tim)
    OSI Addressing (to be determined)
    XNS Addressing (Ed)




Respectfully submitted...

Ed Reed
<[e-mail]>
/c=us/a=telemail/p=xerox-xns/o=xerox/ou=cin ops/cn=eer