Python for large projects

[Peter Hickman]

>>   That is actually a view held by some people and has been attempted on
>>some projects: the programmers do not have access to compilers and deliver
>>their code to someone else when complete.

I seem to recall this, it was a 'zero defect' policy that IBM OZ tried. 
It was one of those PHB development methodologies that assumed that if 
you ran a modern code shop like 'the old days' (of code listings which 
had to be delivered to the punch room to be turned into punch cards that 
were then scheduled to be loaded into the hopper to be compiled so you 
could get a compiler error report two days later - yes thats how we 
actually used to do it!). The assumption was that 'in the old days' the 
code was less buggy thus the methods used must be superiour.

The fault with this was that the amount of code being produced 'in the 
old days' was less, much much less.

>>I think all these projects failed.

Many man years down the drain if I recall, but being IBM they just went 
and tried it all over again.

> Do you have a reference describing such an attempt?

Fraid not, it is folklore for us (ex)IBM mainframe programmers but I do 
recall reading the actual article sometime in the past, oh, 10 years!