Algorithm 版 (精华区)
发信人: ssos (存在与虚无·守拙), 信区: Algorithm
标 题: On Computer System Challenges
发信站: 哈工大紫丁香 (2003年07月27日11:16:15 星期天), 站内信件
On Computer System Challenges
FERNANDO J. CORBAT\'O
MIT
When one considers open problems in Computer Science, it is important
to recognize that, in contrast to the neat welldefined problems of
the theoretical side of Computer Science, Computer Systems deal with
engineering systems that meet societal needs. Solutions often change
with time and frequently require evolution and adaptation. As
computers have become ubiquitous in our society over the last fifty
years, they have become more and more interconnected, culminating in
the Internet and the World Wide Web. Not surprisingly, these changes
have had great social and political impact and seem likely to evolve
at a rapid pace for many decades. Nevertheless, even though today we
are far from a stationary computer and network infrastructure, it is
possible to identify a few fundamental challenges that should be
addressed. All of these problems have been identified before by
others, and indeed there is a great deal of current activity in both
research and development, but I believe it is useful to spotlight
these central challenges.
First, there is a problem of trustworthiness and authentication. Even
within an isolated computer, one cannot be sure that the system
behavior will be correct if the programs and system are not
potentially open for inspection and verification. In particular,
blackbox systems are a major difficulty because they require blind
faith. For example, we see this playing out in criticisms of
electronic voting systems with proprietary software. But
trustworthiness is especially a problem when interacting over the
Internet. Today, every user must make assumptions about the identity
of those he or she is interacting with since the network
infrastructure and protocols in use allow header information to be
forged. One egregious problem arising from the current loose framework
of authentication is the nearexponential rise of spam (i.e., junk
mail). It seems clear that within a few years unless some major
engineering changes are made, the convenient use of electronic mail
will be destroyed by irresponsible spammers drowning out meaningful
communication.
Second, there is the challenge of governance of international
information exchange. Today's Internet is governed in a de facto sense
by the control of the domain name server machinery and by the
agreedupon protocols of network messages. At present, managing the
Internet is like governing without a constitution. This challenge may
appear entirely political, but effective solutions require good
technical support such as the development of the Open Source movement,
the infrastructure of anonymous servers, and cryptographic tools to
avoid tyrannical control over the free flow of information.
A third challenge is the need to create permanent information that is
capable of being retained indefinitely for millennium after
millennium. What makes this problem critical is that most records and
documents today are stored digitally without any paper records. There
are projects addressing this problem but the difficulties are
daunting. Electronic media are fragile: magnetic tapes have lifetimes
measured in decades at most and CDs are only good at best for a few
centuries; there is no equivalent of electronic parchment. But even
more challenging is the continual evolution of technology that
obsoletes the read/write devices for the media. Moreover, the
computing machinery keeps changing along with the computer
languages. In a little more than half a century, we have a sizeable
list of ancient machines, forgotten languages and unreadable
databases. Today, with some of the participants still alive, it is
often possible to resurrect the details of lost machines and
languages, but the difficulty can only increase with time especially
as the complexity of machines and devices continues to grow. And
matters get even more complicated when one tries to keep data
permanently secret by encryption since, on the one hand, keys may be
lost and, on the other hand, bruteforce decoding attacks becomes ever
more feasible.
This list of three challenges is hardly exhaustive, but in focusing on
them, I have tried to address the most serious problems in the area of
computer systems. Optimists may feel that a laissezfaire marketplace
approach will automatically lead to solutions. However, this seems
unlikely since these problems are longrange and do not offer many
opportunities for shortterm commercial gain. Nevertheless, even if
there are no tidy solutions to these challenges and progress must be
evolutionary, it should be valuable for future system designers and
researchers to focus on them as goals.
REFERENCES
For an interesting approach to fight spam see:
FAHLMAN, S. E. 2002. Selling interrupt rights: A way to control
unwanted email and telephone calls. IBM System Journal 41, 4,
pp. 759--766.
For a good survey of data preservation problems see:
TRISTRAM, C. 2002. Data Extinction. Technology Review 105, 8,
pp. 36--42.
--
<<社会契约论>>是一本好书,应当多读几遍
风味的肘子味道不错,我还想再吃它
※ 来源:·哈工大紫丁香 bbs.hit.edu.cn·[FROM: 202.118.230.220]
Powered by KBS BBS 2.0 (http://dev.kcn.cn)
页面执行时间:3.075毫秒