your experience with non-functional requirements

I don't remember ever being asked to develop systems that are non-functional. If I had it would have made my life a lot easier and a considerable saving on the BOM.

Peter

I think he's referring to technical requirements as opposed to functional requirements (in a project management sense of both terms). However the web site to which he pointed is so broken that I can't really be sure.

Hi, that's right, it is about non-functional requirements as they are known in the software-engineering.

Sorry that the server sometimes is not responding, but the link URL is correct.

Julian

It's true that "non-functional requirements" can lead to a non-functional system but the OP's question is about requirements that don't *function* as a requirements should. The litmus test for a functional requirement is Can it be measured? If it can't be measured, you can't tell if the requirement has been met. A more useful way to express a functional requirement is that there can be one and only one interpretation.

Examples:

Non-functional - user friendly Functional - user can edit data value XXX from the root menu with three or fewer button presses

Non-functional - fast response Functional - response screen is painted within XXX milliseconds of external event YYY

Writing useful requirements is a skill that one learns over many years. And I've even seen experts occasionally write non-useful requirements. But that's where the grey area comes in. A useful requirement is as much a function of the local departmental vernacular as it is about the writer's native language. IOW, a useful requirement is useful only as long as everyone who uses it interprets it the same way. The risk is you sometimes never know who will be reading your requirements.

JJS

OTOH, and for example, reducing "user friendliness" to component metrics, then designing to the metrics, can result in a system that's pretty darn user hostile. I don't think that a few such guidelines in a functional specification are off base, as long as they're delineated as such.

The rules are very simple:

1. The customer provides the functional requirements.

1. Engineering generates technical requirements from the functional requirements.

Every link I tried was 404'ing.

As one who has dealt with understanding and writing "Functional Requirements" and "Non-functional Requirements" I can assure you all that they can both be measured for compliance. Larwe mentioned also "Technical Requirements" which could be a mixture of the two.

The Functional Requirements are keyed to the tasks that the equipment must perform and will include event timings, what it must do, how errors are expected to be managed and even the Safety Requirements.

The non-functional requirements have details like how the box in which the equipment is mounted is fixed to the wall or specifies the rack slot dimensions into which it might be plugged. It may even have details of the connectors to be used (pins, style - coded orientation etc) and will also contain the maximum expected weight of the unit and its environmental performance.

Functional Requirements, Non-functional Requirements, Interface Control Document, Safety Requirements and Performance Criteria may all appear in one physical document from which you have to sift each of the appropriate aspects of specification. Then again, they may also appear as separate documents in their own right.

PS. While the webpage may be colourful it is not doing very much on my browser (a list of a few disparate items and a panel stating that it is "Loading").

On a multichannel spectrum analyzer I had a requirement that the signal processing be performed in a floating point format of at least

This was verified by inspection of code and compiler manuals rather than a functional test.

Dale B. Dalrymple

I think a few people missed the joke... ;)

Regards,

Indeed, or have an even drier sense of humour than I.

Peter

That's what the smiley was invented for, 25 years ago...

How does one categorize requirements such as "It has to use my cousin's company's design software" or "I heard fuzzy logic is good. Use some of that. And maybe some neural nets" ?

A non-functional requirement. The question when responding to this is whether or not the cousin's company's design software would allow you to meet the goals of the project or not.

nets" ? In a requirements specification it is borderline between functonal and non-functional. However, this is specifying implementation details too early really and the question here is whether or not you might specify fuzzy logic or neural nets in the Technical Specification (a much more detailed level).

I thought that if the customer wants something to be used than it is called a constraint. It's no more a requirement, neither functional nor non-functional. If it makes sense is then another question.

To all users in this discussion: thank you. I just wanted to say, that I'm still following the discussion. Concerning the above mentioned web-application: I'm about to move it on the server of the science chair where I write my diploma thesis. I hope their server runs more stable than the one I'm using right now. The URL will be given in time.

Using some fuzzy logic is easy:

2) Be very fuzzy in documenting how it works. If pressed for details, point out that they specifically wanted a fuzzy solution!

Hi all, the webapplication for sharing your knowledge has finally moved. The new URL is

Thank you, J. Wild