rfc1210.txt

著名的RFC文档,其中有一些文档是已经翻译成中文的的.

   complete environment.  Several problems have to be resolved to
   appropriately handle this situation.  The first problem is the
   global-naming of bitfiles that are being moved through the DCE
   to/from the archive.  Second, the file system hierarchy must be
   defined.  Third, there is the question of how the DCE knows the file
   system hierarchy for which it is responsible, and the location of the
   boundary through which the network and the archival system operate.
   Lastly, there is the question how the file system hierarchy is
   divided across a DCE and within a supercomputer.

   A second issue in the DCE is the need for all nodes obtaining or
   storing data to know the storage media differences.  For future
   systems, this requirement manifests itself both at the distributed
   nodes and at the supercomputer because of the differences in the
   physical media structure.

   The third issue is the delineation of the bitfile attributes.  This
   relates to how the data must be maintained as it migrates through the
   hierarchy, as well as through the DCE.  The bitfile carries
   attributes based upon its location in the hierarchy, or in the DCE,
   that may be different from those needed at the supercomputer level.
   Many of these attributes are related to the data content and where it
   resides in time within the DCE.  Section 6.15 discusses some of the
   possible meta-data representation methodologies that may be used but
   are not yet standardized.

   Another issue is the determination and implementation of the site
   policy that is to dictate data migration and allocation inside the
   DCE archival storage system.

   Several working committees are attacking the various problems
   delineated above, and are trying to confront the difficulties in
   these environments.  This work is progressing mostly in the United
   States.  The IEEE Computer Society Technical Committee on Mass



Cerf, Kirstein, & Randell                                      [Page 22]

RFC 1210      Network and Infrastructure User Requirements    March 1991


   Storage Systems is in the process of developing a Computer Society
   draft standard on data storage systems.  The current working draft
   provides a consistent terminology and an object-oriented design for
   defining storage subsystem components, whether they are being built
   around a single system or in a DCE.  Other groups in the computing
   community are currently dealing with the problems of data migration
   within a distributed environment.  This distributed environment may
   or may not include a supercomputer, but it almost always includes a
   high-volume storage system of some sort delineated as a "mass storage
   system." This subject was not discussed long enough at the meeting to
   deduce one year or three year targets - indeed these may well be set
   by the relevant National working groups.

6.14.1  Recommended Actions

   Convene an international workshop whose goals are:

   1.  An understanding of the contents of the data storage reference
       model that is nearly ready to be declared an official standard
       guide;

   2.  To continue discussion of the various system issues that have
       to be developed as a result of this model;

   3.  To arrive at solutions to be proposed by vendors and users for
       implementations of Data Systems Storage Solutions which are
       modular, interconnectable, and standard.

6.15  Data Representation and Exchange

   The problem of data exchange between different computer architectures
   and operating systems has been existent since the deployment of the
   early computers.  This problem has been exacerbated by the acceptance
   of the client-server paradigm as the provider of distributed
   services.  Distributed computer services require immediate data
   exchange.  In the past, data was exchanged on some medium, such as
   tape, and could be examined at leisure.  Ad hoc data conversion
   routines were created to process the data, and were often embedded in
   the programs using the data.  Data exchange in the client-server
   paradigm does not permit this leisurely data examination.  Both the
   client and the server must be able to "call" software that is
   guaranteed to convert the exchanged data "on the spot."  This
   guarantee also implies a standard format rather than the ability to
   convert all formats because it would be impossible to maintain
   multiple architecture conversion software and, of course, the size of
   such conversion software would be enormous.

   The issue of data exchange has been addressed resulting in many data



Cerf, Kirstein, & Randell                                      [Page 23]

RFC 1210      Network and Infrastructure User Requirements    March 1991


   exchange software packages.  A few of the currently more popular
   packages are XDR, HDF, NetCDF, PostScript and CCSDS.  Each of these
   packages addresses a specific type of data.  Some address bitmap
   data; one addresses the general encoding of "display" information.
   Some of these packages address various numerical representations in
   computers.  It is unclear whether any existing package could or
   should be extended to solve all data exchange problems.  However, a
   more realistic approach would be a collection of data exchange
   packages formed as the "standard."

   This item was discussed only briefly at the meeting, so that no one
   year or three year targets were specified.

6.15.1  Recommended Actions

   It is proposed that an international working group be established to
   recommend a standard collection of software encompassing a variety of
   data representations.  This working group should address the issue of
   embedding identification of the data representations in the data
   stream to allow for later extensions.  The working group would meet
   initially to establish a work-scope and to assign the members tasks.
   The group would schedule subsequent meetings (probably annually) to
   finalise the current data exchange standard recommendation, and to
   define new work scopes.  The working group would also make their
   recommendation known to other standards bodies such as X/OPEN, UI,
   OSF, X Consortium, NIST, IEEE, ACM, etc.

6.16  Transatlantic Links and Continental Distribution

   At present, there is inadequate transatlantic capacity to support
   research collaborations involving significant amounts of computer-
   mediated communication.  There is also considerable room for
   improvement in the distribution of capacity and enhancement of
   reliability of network service in Europe.  Moreover, the point was
   made strongly that collaboration would be very difficult unless the
   infrastructure on the two sides was broadly comparable - even if the
   transatlantic capacity was per force lower.  Moreover, it was sharply
   emphasised that there was a large requirement for transatlantic data
   flow in other fields - e.g., Space Science, Atmospheric Science and
   High Energy Physics.  In the US these needs are being aggregated in
   the National Research and Engineering Network; such aggregation is
   required also in Europe and on a transatlantic basis.

6.16.1  One Year Targets

   (i)  Install 2 Mb/s multi-protocol distribution facilities in Europe;

   (ii) Install 1.5 Mb/s (or higher) transatlantic capacity.



Cerf, Kirstein, & Randell                                      [Page 24]

RFC 1210      Network and Infrastructure User Requirements    March 1991


6.16.2  Three Year Targets

   (i)  Install 2 additional 1.5 Mb/s (or higher) transatlantic links
        by 1993;

   (ii) Determine feasibility of sharing much higher bandwidth
        intercontinental links (e.g., in the 51 Mb/s STS-1 range).

6.16.3  Recommended Actions

   (i)   Use existing joint US/European coordination mechanisms
         (e.g., CCIRN) for planning of higher speed, transatlantic
         links;

   (ii)  Convene a special CEC/DARPA/NSF task force to consider much
         higher speed transatlantic capacity sharing options;

   (iii) Ensure that there is an infrastructure in Europe, paralleling
         the US one, providing the majority of relevant campuses access
         at speeds approaching 1.5 Mb/s;

   (iv)  Encourage European user groups with high data transmission
         requirements to aggregate their data transmission facilities.
         Attempt to integrate European application projects (like the
         RACE Applications Pilots) to assist in providing an appropriate
         European distribution network with 10 - 500 Mb/s access to
         appropriate campuses.

7.  LONGER TERM INITIATIVES

   Although these were not discussed in any detail, for lack of time,
   the following areas emerged as of interest for longer term
   collaborative work:

   (i)   Electronic Library Services (includes an important
         intellectual property rights component);

   (ii)  Multi-media Computer Supported Collaborative Work;

   (iii) Portable Computing/Communications Environments;

   (iv)  Distributed Computing using heterogeneous machines and unique
         facilities;

   (v)   Compatible approaches to computer networks with Gb/s access
         speeds, and appropriate systems switching, transmission and
         protocols.




Cerf, Kirstein, & Randell                                      [Page 25]

RFC 1210      Network and Infrastructure User Requirements    March 1991


   It was felt that some collaborative research in these areas would
   have immense medium term benefits to the other communities - and
   would integrate well with the ongoing research programmes on both
   sides of the Atlantic.

8.  OBSTACLES

   The largest single obstacle to the provision of the facilities
   outlined in this report are that development of the necessary
   facilities do not have priority to most funding agencies.  This is
   exemplified by the role of our workshops in this series.  Not only
   network provision, but also development of appropriate infrastructure
   application software and testbed activity, are essential.

   There are a number of problem areas which could benefit from official
   attention from CEC and US research funding agencies.  For example,
   there are a number of open and proprietary protocol suites which are
   candidates for use in US/European collaborative research.  However,
   there is lack of political agreement as to how to deal with these
   various suites.  It would be politically valuable if the CEC and US
   research agencies could issue a communique outlining common agreement
   on treatment of multiple protocols (e.g., expressing serious interest
   in supporting campus-to-campus communication using multiple
   protocols).  Within the OSI protocol suite, there are differences as
   to which features ought to make up the standard profile for use by
   government-sponsored groups.  Handling of connection-oriented and
   connectionless protocol elements within the suite is the subject of
   continued debate.  Agreement to support at least TCP/IP and the
   connectionless network protocol in the OSI suite on an
   intercontinental basis would be beneficial to both parties; many CEC
   members would like connection-oriented network protocols to be
   supported also.

   European international tariffs are relatively high.  This has
   inhibited the implementation of private networks and impeded progress
   on collaborative work between the US and Europe.  A CEC initiative to
   come to grips with this problem could be quite helpful.

   There are a diversity of intra-European networking organizations
   which have technical, operational and policy interests.  Planning for
   intercontinental networking infrastructure is sometimes confused by
   the variety of interested parties.  Effort towards further
   coordination and rationalization of intra-European networking
   activities could make intercontinental planning somewhat easier.

   There is a strong interest in the use of cryptographic methods to
   provide privacy, authenticity and integrity assurance for various
   forms of intercontinental communication and at various levels in the



Cerf, Kirstein, & Randell                                      [Page 26]

RFC 1210      Network and Infrastructure User Requirements    March 1991


   protocol hierarchies.  Although there appears to be substantial
   technical activity in this area, progress is now impeded by national
   restrictions on the export of software which utilizes cryptographic
   methods.  National use policies vary and the ability to apply these
   valuable and needed techniques is uncertain.

   Some national privacy and data protection laws prohibit the creation
   of directories containing personal information (e.g., email and
   postal addresses) and other laws limit what kinds of information (and
   in what form) can be transported across national borders.

   Handling of cryptographic exchanges, import/export of supporting
   software and exchanges of keying information are all potentially
   subject to national restrictions and constraints.  The government
   agencies interested in promoting international collaboration may need
   to seek alternative international formulations of permitted practice
   to permit the required technical support.

   Finally, several organizations in the US and Europe have pointed out
   that the provision of networking infrastructure requires stable
   funding over significant periods of time.  Stability for
   infrastructure support has been shaky in the US and in Europe and
   this presents an obstacle to achieving widespread and reliable
   network services to aid collaborative efforts.

9.  CONCLUDING REMARKS

   The set of proposals contained in this report provide a realistic,
   staged approach to