RFC du protocole NNTP : introduction
1. Introduction
For many years, the ARPA-Internet community has supported the
distribution of bulletins, information, and data in a timely fashion
to thousands of participants. We collectively refer to such items of
information as "news". Such news provides for the rapid
dissemination of items of interest such as software bug fixes, new
product reviews, technical tips, and programming pointers, as well as
rapid-fire discussions of matters of concern to the working computer
professional. News is very popular among its readers.
There are popularly two methods of distributing such news: the
Internet method of direct mailing, and the USENET news system.
1.1. Internet Mailing Lists
The Internet community distributes news by the use of mailing lists.
These are lists of subscriber's mailbox addresses and remailing
sublists of all intended recipients. These mailing lists operate by
remailing a copy of the information to be distributed to each
subscriber on the mailing list. Such remailing is inefficient when a
mailing list grows beyond a dozen or so people, since sending a
separate copy to each of the subscribers occupies large quantities of
network bandwidth, CPU resources, and significant amounts of disk
storage at the destination host. There is also a significant problem
in maintenance of the list itself: as subscribers move from one job
to another; as new subscribers join and old ones leave; and as hosts
come in and out of service.
1.2. The USENET News System
Clearly, a worthwhile reduction of the amount of these resources used
can be achieved if articles are stored in a central database on the
receiving host instead of in each subscriber's mailbox. The USENET
news system provides a method of doing just this. There is a central
repository of the news articles in one place (customarily a spool
directory of some sort), and a set of programs that allow a
subscriber to select those items he wishes to read. Indexing,
cross-referencing, and expiration of aged messages are also provided.
1.3. Central Storage of News
For clusters of hosts connected together by fast local area networks
(such as Ethernet), it makes even more sense to consolidate news
distribution onto one (or a very few) hosts, and to allow access to
these news articles using a server and client model. Subscribers may
then request only the articles they wish to see, without having to
wastefully duplicate the storage of a copy of each item on each host.
1.4. A Central News Server
A way to achieve these economies is to have a central computer system
that can provide news service to the other systems on the local area
network. Such a server would manage the collection of news articles
and index files, with each person who desires to read news bulletins
doing so over the LAN. For a large cluster of computer systems, the
savings in total disk space is clearly worthwhile. Also, this allows
workstations with limited disk storage space to participate in the
news without incoming items consuming oppressive amounts of the
workstation's disk storage.
We have heard rumors of somewhat successful attempts to provide
centralized news service using IBIS and other shared or distributed
file systems. While it is possible that such a distributed file
system implementation might work well with a group of similar
computers running nearly identical operating systems, such a scheme
is not general enough to offer service to a wide range of client
systems, especially when many diverse operating systems may be in use
among a group of clients. There are few (if any) shared or networked
file systems that can offer the generality of service that stream
connections using Internet TCP provide, particularly when a wide
range of host hardware and operating systems are considered.
NNTP specifies a protocol for the distribution, inquiry, retrieval,
and posting of news articles using a reliable stream (such as TCP)
server-client model. NNTP is designed so that news articles need only
be stored on one (presumably central) host, and subscribers on other
hosts attached to the LAN may read news articles using stream
connections to the news host.
NNTP is modelled upon the news article specifications in RFC 850,
which describes the USENET news system. However, NNTP makes few
demands upon the structure, content, or storage of news articles, and
thus we believe it easily can be adapted to other non-USENET news
systems.
Typically, the NNTP server runs as a background process on one host,
and would accept connections from other hosts on the LAN. This works
well when there are a number of small computer systems (such as
workstations, with only one or at most a few users each), and a large
central server.
1.5. Intermediate News Servers
For clusters of machines with many users (as might be the case in a
university or large industrial environment), an intermediate server
might be used. This intermediate or "slave" server runs on each
computer system, and is responsible for mediating news reading
requests and performing local caching of recently-retrieved news
articles.
Typically, a client attempting to obtain news service would first
attempt to connect to the news service port on the local machine. If
this attempt were unsuccessful, indicating a failed server, an
installation might choose to either deny news access, or to permit
connection to the central "master" news server.
For workstations or other small systems, direct connection to the
master server would probably be the normal manner of operation.
This specification does not cover the operation of slave NNTP
servers. We merely suggest that slave servers are a logical addition
to NNTP server usage which would enhance operation on large local
area networks.
1.6. News Distribution
NNTP has commands which provide a straightforward method of
exchanging articles between cooperating hosts. Hosts which are well
connected on a local area or other fast network and who wish to
actually obtain copies of news articles for local storage might well
find NNTP to be a more efficient way to distribute news than more
traditional transfer methods (such as UUCP).
In the traditional method of distributing news articles, news is
propagated from host to host by flooding - that is, each host will
send all its new news articles on to each host that it feeds. These
hosts will then in turn send these new articles on to other hosts
that they feed. Clearly, sending articles that a host already has
obtained a copy of from another feed (many hosts that receive news
are redundantly fed) again is a waste of time and communications
resources, but for transport mechanisms that are single-transaction
based rather than interactive (such as UUCP in the UNIX-world <1>),
distribution time is diminished by sending all articles and having
the receiving host simply discard the duplicates. This is an
especially true when communications sessions are limited to once a
day.
Using NNTP, hosts exchanging news articles have an interactive
mechanism for deciding which articles are to be transmitted. A host
desiring new news, or which has new news to send, will typically
contact one or more of its neighbors using NNTP. First it will
inquire if any new news groups have been created on the serving host
by means of the NEWGROUPS command. If so, and those are appropriate
or desired (as established by local site-dependent rules), those new
newsgroups can be created.
The client host will then inquire as to which new articles have
arrived in all or some of the newsgroups that it desires to receive,
using the NEWNEWS command. It will receive a list of new articles
from the server, and can request transmission of those articles that
it desires and does not already have.
Finally, the client can advise the server of those new articles which
the client has recently received. The server will indicate those
articles that it has already obtained copies of, and which articles
should be sent to add to its collection.
In this manner, only those articles which are not duplicates and
which are desired are transferred.
|