Re: [SystemSafety] Does "reliable" mean "safe" and or "secure" or neither?

From: Mike Ellims < >
Date: Thu, 28 Apr 2016 13:51:13 +0100


Good afternoon Michael,

Well that's a bit of a damp squib then, I was expecting a fiery return comment!

Cheers.

-----Original Message-----
From: systemsafety
Michael J. Pont
Sent: 27 April 2016 18:41
To: 'The System Safety List'
Subject: Re: [SystemSafety] Does "reliable" mean "safe" and or "secure" or neither?

Mike (Ellims),

Thanks for your reply.

My original email was intended to make two points, the first of which C. Michael Holloway expressed much more clearly and succinctly than I did in a later post:

"One would surely expect professional engineers to be less cavalier in frequently adding new definitions to common words, which change the meanings in confusing, unhelpful ways."

I (still) believe that this is what has happened here in the application of the phrase reliability to software.

I'm happy to accept your argument that - when following particular standards - we may be obliged to accept the definitions that are associated with these standards (but my original point was intended to be more general).

Michael.

-----Original Message-----
From: Mike Ellims [mailto:michael.ellims_at_xxxxxx Sent: 25 April 2016 13:25
To: M.Pont_at_xxxxxx <systemsafety_at_xxxxxx Subject: RE: [SystemSafety] Does "reliable" mean "safe" and or "secure" or neither?

Good afternoon Michael,

First sorry about the length of the reply and the delay!

>>
>> We can choose to define any terms that we use in a particular domain
>> in

any way that we like.
>>

First who is we?

Second, no we can't; to maintain consistency with existing standards, practice and research we have to maintain constancy with existing definitions, for example IEC 61508 Part 4 states "see IEV 191-12-01 for a definition of reliability".

IEV 191-12-01 defines reliability as, "probability that an item will perform a required function under given conditions for a given time interval (t1.t2)". ISO 26262 doesn't give a definition of reliability but as it purports to be a sector specific derivative of IEC 61508 I assume it will inherit definitions not otherwise defined (does anyone have an answer to that?).

>>
>> Where these terms conflict with general use of the same word, we make
>> it

more likely that people reading the
>> documents - etc - will be confused (in my view).
>>

Which is why standard definitions such as that given in BS 4478 and IEV 191-12-01 exist because there is a disparity between what the man in the street means when they say reliable and what reliability engineers have decided that term reliable means. The same problem exists with the use of the word theory, in science is it a reasonably precise meaning and scientists have other words such as conjecture and hypothesis for concepts that are not well established. However in common use the word theory has a different less precise meaning and in general represents a plausible scenario as may be argued in court of law.

>>
>>Confusion tends to result in Bad Things Happening in this business.
>>We

(therefore) want to avoid confusion.
>>

Which is why we have an exist precise definition... actually I'll concede that as stated it's not actually very precise i.e. it doesnít specify how to perform the statistical evaluation buy thatís expanded on in textbooks such as O'Conner and IEC 61508 etc. Note this is not an argument for the usefulness of the term as applied for software it's only an argument against redefining the meaning of the term. In general I would argue that if we have a problem with the way a term is defined then we should find a similar term that could be explicitly defined to mean what we think is necessary. So rather than try and redefine "software reliability" or "reliability" we should use a different term.

>>
>> Two seconds on Google gives me this definition of reliability:
>>

Sorry but I don't think that appeal to Google is a reasonable approach. Google can be used as a mechanism for locating specific information in a field of study e.g. reliability engineering. For example if I put the search term "reliability" into Google the first entry is a link to the Wikipedia disambiguation page that lists a set specialist used for the term reliability (https://en.wikipedia.org/wiki/Reliability) the second of which is reliability engineering. The second sentence of the entry for reliability engineering gives the following definition "Dependability, or reliability, describes the ability of a system or component to function under stated conditions for a specified period of time" which is cross referenced to IEEE Standard Computer Dictionary: A Compilation of IEEE Standard Computer Glossaries.

Various generic dictionaries produce the following set of definitions in order of appearance;

dictionary.com : the ability to be relied on or depended on, as for accuracy, honesty, or achievement.
merriam-webster.com : 1. the quality or state of being reliable 2: the extent to which an experiment, test,

                          or measuring procedure yields the same results on
repeated trials
thefreedictionary.com : The ability of an item to perform a required function under stated conditions for a
                          specified period of time.
en.wiktionary.org       : 1. The quality of being reliable, dependable or
trustworthy. 2. In education - the ability to
                          measure the same thing consistently (of a
measurement indicating the degree to which the
                          measure is consistent); that is, repeated
measurements would give the same result (See also
                          validity). 3. In engineering - measurable time of
work before failure

Several definitions are a close match the IEC and IEEE definition and the last makes a clear distinction as to where different definition apply. However there is wide variation and taking this approach it's pot-luck whether any particular person would derive anything useful or usable. For example the Concise OED has no separate definition for reliability and the Oxford Dictionary of Computing does and it matches the IEC/IEE/BS standard.

Note that if we search for the term "software reliability" we usually get something more akin the standard meaning as used in reliability engineering e.g. "A software quality aspect that is measured in terms of mean time to failure or failure intensity of the software", or from ANSI " the probability of failure-free software operation for a specified period of time in a specified environment".

>> "Definition of reliability. 1 : the quality or state of being
>> reliable. 2

:
>> the extent to which an experiment, test, or measuring procedure
>> yields

the same results on repeated trials."
>>
>> I see "2" as the useful (and general) definition here. It may even
>> be

compatible with O'Conner.

I don't think that this is compatible with O-Conner nor the standard definitions of engineering reliability, and to my mind the second definition would seem to be a closer match to testing or program corretness.

As an extreme contrived counter example to the usefulness of definition 2 we could consider the case of a software system that always halted 1/2 a second after start producing no output. We could run this any number of times and get the same output (whatever the input) and meet the criteria above that repeated trails produced the same output. Thus the software system is reliable. Useless but reliably useless... which is perhaps not useful.

>>
>> My issue is that software (whether good or bad) doesn't change over time.
>>

This statement uses the starting position that all reliability issues are a result of wear out, i.e. X changes over time.

My original point was that reliability engineering is NOT just about wear out. An object (mechanical, electric, software) can fail to deliver the specified function for a number of different reasons. A simplified mechanical example may help, consider if a part were made from aluminum as specified, rather than stainless steel as was intended (i.e. we have a specification error) then the part could fail more quickly than desired for several reasons;

    i) it was not strong enough,
    ii) it melted,
    iii) it suffered badly from corrosion,     iv) or because it wore more quickly.

Reliability isn't only about "wear-out"; as defined it's "about failure to deliver functions", whether that function be ability of hold a load, support a beam, maintain a current, measure a value, monitor a parameter or to perform a calculation. In addition you can stretch the analogy of wear out to żNiklaus Wirth's concept of accumulated state. That is the longer the program runs the more (possibly redundant) state information builds up within in memory, the more probable the system is to go bonk (important technical term). The patriot missile system springs to mind as a simple example of catastrophic state accumulation and while possibly reliable when used with the assumed timeframe (t1,t2) it proved unreliable the used in the timeframe (t1,t3) where t3 >> t2.

>>
>> Talking about "software reliability" (therefore) doesn't make sense
>> and

is likely to lead to confusion.
>> (Hardware reliability is fine; System reliability is also fine, where
System = Hardware + Software.)
>>

At the current time, given the limits associated with research in software reliability prediction that is not that an unreasonable position to take in practical terms. However as stated above I don't think that gives a green light to appropriate the term reliability.

As a topic of theoretic interest software reliability is a valid area of research and some of the concepts have at least some utility if the limitations of the techniques are taken into account. For example reliability growth theory indicates that over time with a good development process the rate of failures observed should decline; so it isn't then we are in an effluent stream with no propulsive device (grossly simplified of course). There is also practical application to the evaluation of systems fielded and to comparisons between different software running on identical platforms.

In summary:
1. Reliability is a existing defined term, even for software. 2. For software the term may have little practical predictive power, however it can't just be redefined.
3. If necessary a new term should be defined that captures the idea of reliability in a manner that matches common usage. 4. I suspect 3 could be really, really, really hard.

Please feel free to take pot-shots.

Cheers.

-----Original Message-----
From: systemsafety
[mailto:systemsafety-bounces_at_xxxxxx Mike Ellims
Sent: 24 April 2016 13:23
To: 'Coq, Thierry' <Thierry.Coq_at_xxxxxx <systemsafety_at_xxxxxx Subject: Re: [SystemSafety] Does "reliable" mean "safe" and or "secure" or neither?

> as software does not "fatigue" or randomly "break" the way hardware > does

Reliability engineering is much more than just fatigue or randomly breaks, it encompasses everything about the control of variation and error in a system, most commonly using statistical methods.

BS 4478 defines reliability as "The ability of an item to perform a required function under stated conditions for a stated period of time".

O'Conner states that "reliability is usually concerned with failures in the time domain. This distinctions marks the difference between traditional quality control and reliability engineering".

He goes on to list a number of reasons why a failure may occur as follows (abridged)

Henley and Kumamoto give a potted history of the development of reliability engineering and track its roots to work done by Lusser on the V2 (V1?) missile which was spectacularly unreliable at first. But obviously not a wear out issue...

This is of course separate from whether in any context reliability is useful concept e.g. software. You can obviously measure the reliability of Windows98 is terms of mean time between crashes (failures in time) likewise you can measure the reliability of Linux in the same manner. Whether that has any deep meaning aside from the fact that it shows Windows98 to be pants compared to Linux for some distribution of uses/input/outputs a different question.

Cheers.

-----Original Message-----
From: systemsafety
Coq, Thierry
Sent: 24 April 2016 07:24
To: The System Safety List
Subject: Re: [SystemSafety] Does "reliable" mean "safe" and or "secure" or neither?

Hi
I find this list hugely informative. In particular, I find PBL's posts factual, useful, interesting and intriguing. As well as the many other debaters who agree with him or challenge him. As it should be. I wish to express my gratitude to all debaters.

However, this last exchange seems to me a debate on authority. On our left, we have DO-178. B now C.
On our right, we have IEC, IEEE, Musa, etc.

To go further, it is plain fact that the aeronautics industry has demonstrated it doesn't need "software reliability" to deliver highly reliable automated systems, or systems of systems. It seems evident with the knowledge we have of the aeronautics success that in order to use "software reliability" in other industries, or in aeronautics, there needs to be a clear use case where the value of "software reliability" is demonstrated, compared to other methods or techniques, in order to apply "software reliability". The analogy of software and hardware does not seem valid, as software does not "fatigue" or randomly "break" the way hardware does, which is the basis for all probabilistic reliability theories for hardware. The analogy that does seem valid between hardware and software is the presence of systematic faults, in design, manufacturing, installation, testing, misuse, etc. Which in hardware also does not have a probability number. Formal methods can be used to identify such systematic faults in software. If one can be found, then a test environment can be devised in which the software will fail 100% of the time. Random hardware faults do not behave like t hat.

Best regards,
Thierry Coq
The opinions reflected here are my own and are not necessarily those of my employer

-----Original Message-----
From: systemsafety
[mailto:systemsafety-bounces_at_xxxxxx Peter Bernard Ladkin
Sent: dimanche 24 avril 2016 03:27
To: The System Safety List
Subject: Re: [SystemSafety] Does "reliable" mean "safe" and or "secure" or neither?

On 2016-04-23 19:43 , Nick Tudor wrote:
> DO-178C

In the absence of a complete sentence, let me suggest one.

It's fine for some industry consortium to find it has no use for a specific concept. RTCA likely has no use for the notion of a cup of tea, either (BS6008). But that doesn't mean it makes any sense to argue that there isn't any such thing as a cup of tea.

PBL Prof. Peter Bernard Ladkin, Faculty of Technology, University of Bielefeld, 33594 Bielefeld, Germany Je suis Charlie Tel+msg +49 (0)521 880 7319 www.rvs.uni-bielefeld.de




This e-mail and any attachments thereto may contain confidential information and/or information protected by intellectual property rights for the exclusive attention of the intended addressees named above. If you have received this transmission in error, please immediately notify the sender by return e-mail and delete this message and its attachments. Unauthorized use, copying or further full or partial distribution of this e-mail or its contents is prohibited.



The System Safety Mailing List
systemsafety_at_xxxxxx
---
This email has been checked for viruses by Avast antivirus software.
https://www.avast.com/antivirus

_______________________________________________
The System Safety Mailing List
systemsafety_at_xxxxxx

_______________________________________________
The System Safety Mailing List
systemsafety_at_xxxxxx


---
This email has been checked for viruses by Avast antivirus software.
https://www.avast.com/antivirus


_______________________________________________
The System Safety Mailing List
systemsafety_at_xxxxxx


---
This email has been checked for viruses by Avast antivirus software.
https://www.avast.com/antivirus

_______________________________________________
The System Safety Mailing List
systemsafety_at_xxxxxx
Received on Thu Apr 28 2016 - 14:51:32 CEST

This archive was generated by hypermail 2.3.0 : Sat Feb 16 2019 - 07:17:08 CET