This page is to refute the idea that XP can be used for ''anything''.  Consider ExtremeSurgery (''aka EmergencySurgery'').  

The Extreme practices tend to be "plan as you go", specifically avoiding BigDesignUpFront.  BigDesignUpFront is, of course, a major value in surgery, since design time is much less expensive (and, more to the point, less dangerous) than time in the OR.  Imagine performing surgery using UnitTest''''''s, RefactorMercilessly,  and YouArentGonnaNeedIt..

	* UnitTest''''''s: ''TestEverythingThatCouldPossiblyBreak because when that patient's heart stops, you want to know!''
	* RefactorMercilessly: ''???''
	* YouArentGonnaNeedIt:  ''The HippocraticOath, as well as the physical limits of the patient's constitution, prevent you from doing things the patient doesn't really need.''
		* ''I thought this was referring to appendectomies and liposuction.  :)''

Programmers boldly go where no one has gone before, and things often blow up in their faces.  These are called "learning experiences".  Surgeons timidly go where others have gone before, and ''very rarely'' do things blow up in their faces.  When they do, they are called "death certificates" and "lawsuits".

Let the programmers program, and the surgeons do surgery.

Just like XP does surgery poorly, it does certain types of programming poorly.  Consider aircraft fly-by-wire code, or other LifeCritical applications.  These require better quality than ExtremeProgramming, or any commercial software methodology, can produce.  These are produced using stricter methodologies, a sort of surgical programming.  The error rate is incredibly low, but the code is incredibly expensive per line of code.

''You've asserted that no commercial software methodology can produce quality sufficient for life-critical software.  So just what methodology '''do''' you use to produce life-critical software?''

----

XP is inherently limited. Refactoring by definition is an adjustment of factors within their context; one conceptual level. Thus it does not reach other conceptual levels effectively, and is not efficient for developing complex systems.

Better methods would involve the establishment of SoftwareArchitecture before a rapid design/ coding process. 'Rapid' approaches could be applied to architecture but in this domain code is not the expressive language and gung-ho coding can not be a relevant behaviour... :-)

It is harder to think than to type.

Life-critical software can be methodologized by *any* methodology which covers the entire stack from customer requirements to general software requirements (in architecture) to design to implementation to testing; establishes both pre- and post-condition testing at every level; and is verified over the entire methodology's use to ensure reliable propagation of pre- and post- tests down the stack.

ThomasWhitmore.

----