defect / fault / failure [LWN.net]

defect / fault / failure

Posted Jul 10, 2008 2:46 UTC (Thu) by zooko (subscriber, #2589)
In reply to: Issue by ncm
Parent article: Dan Kaminsky Discovers Fundamental Issue In DNS: Massive Multivendor Patch Released (Securosis.com)

If we're talking about a problem that exists in the design of DNS or
in a particular implementation, we should call it a "defect" (this is
the standard term among the safety engineers for what we programmers
informally call a "bug").  If we're talking about a problem that
exists at run-time, when something goes wrong internally, we should
call it a "fault".  If we're talking about somebody getting ripped off
because they relied on DNS and a criminal exploited DNS, then we
should call it a "failure".

It is useful to make these distinctions among different kinds of
problems, and using the word "defect" instead of "bug" makes it easier
to communicate accurately with safety engineers from other
disciplines.  (Also, as Eric Hughes suggested to me many years ago,
there may be a psychological benefit of this terminology -- a "bug"
sounds like something endogeneous, but a "defect" is clearly the
responsibility of the engineer who designed the system.)


defect -- A flaw in a system or system component that has the potential to cause that system
or component to fail to perform its required function during execution, [Jones - IBM].  	(
Reference : SEI:SE-CMM)

fault -- an incorrect step, process or data definition in a computer program  	( Reference :
ISO/IEC JTC1/SC7:14598-1)

failure -- The inability of a system or component to perform its required functions within
specified performance requirements. [IEEE STD 610.12-1990]  	( Reference : SEI:SE-CMM)

references:

http://citeseerx.ist.psu.edu/viewdoc/download;jsessionid=...

http://www.totalmetrics.com/cms/servlet/main?Subject=List...

http://en.wikipedia.org/wiki/Safety_engineering

IEEE 610.12-1990, "IEEE Standard Computer Dictionary: A Compilation of IEEE Standard Computer
Glossaries"

(Log in to post comments)

defect / fault / failure

Posted Jul 11, 2008 16:45 UTC (Fri) by giraffedata (subscriber, #1954) [Link]

People like to use the term "bug" for its connotation even when its denotation doesn't apply. "Bug" connotes something bad; something that should embarass someone; something that should be fixed. But its traditional denotation is a programming defect, as distinct from a higher level design defect.

Nobody's saying what this DNS issue is, so I don't know if it's a bug or not, but it sounds to me like the code is all functioning as its designers meant it to, and just not meeting the requirements of today's deployments. I would not use the word "bug" myself.

Also, I'm not sure from what I've read that it's universally regarded as a defect. Are there people defending the existing function? If so, it's more objective to call it an issue than a defect.

An issue is not a euphemism for bug, default, fault, etc. An issue is something people are thinking about.

defect / fault / failure

Posted Jul 11, 2008 17:01 UTC (Fri) by giraffedata (subscriber, #1954) [Link]

fault -- an incorrect step, process or data definition in a computer program ( Reference : ISO/IEC JTC1/SC7:14598-1)

This is a pretty bad definition; it doesn't capture the essential difference between a defect and a fault as the terms are commonly used and as used in the other references.

There's a simple distinction: a defect is a state and a fault is an event. If my email sending program doesn't check the divisor for 0 before dividing by it, that's a defect in the program. Each time I run the program and it crashes because it divides by zero, that's a fault in my sending of the email. If I don't have some way to recover and get the email out anyway, it's also a failure in my sending of the email.