Fog Creek Software
Discussion Board

The real costs of software rewrites

Does anyone have any hard data on how much it actually costs to rewrite an existing piece of software?  I love Joel's article, but it doesn't have any hard data on the subject.


Keith Sader
Wednesday, October 31, 2001

Well, it depends.

What software? How big? What platforms does it need to run on? Does it need a complete rewrite? How many people worked on it the first time? Are you going to be using a different OS/database/programming language, etc.?

The list of questions goes on.

No "one size fits all" formula at all, I'm afraid.

Madhu Menon
Wednesday, October 31, 2001

In my experience deep refraction - you do not change what the thing is doing but you move a lot of code around to provide meaningful structure and do some substantial clean-ups - should not take more that - 10-15% of developing the whole thing from scratch.

Tuesday, November 6, 2001

If you just reimplement the same functionality with higher quality, it generally takes about half the time of the original. If you are adding significant functionality, then you have all the standard schedule-making and -keeping problems.

Later, there will be the shakedown period. If you've documented and designed properly (and not added significant new features) it should be much shorter than the original shakedown period.

The other question to consider is: how much time am I spending debugging/patching this code, including all costs of customer support, industry reputation, etc.

(This is based on my experiences...)

Jeff Kotula
Tuesday, November 6, 2001

*  Recent Topics

*  Fog Creek Home