Originally posted by Kickaha
I have a sneaking suspicion that the article is misinterpreting reality.
60% of Vista may need work. That I completely believe.
60% of 10-15 million lines of code are not going to be rewritten, tested, and approved from scratch in 9 months.
Eight years ago at Nortel we had a project that tried to do that (total re-write of 30 million lines of code) - in the end they threw away the re-written code and went back to the original after wasting a few hundred million dollars.
In general, re-writes are a good thing - as long as you restrict it to a small area at a time, focusing on the sections of code that generate the most problem reports.
If they have this much code that needs to be fixed, then that points to a process problem - you can only get into that situation if you have a lot of "shoot the design from the hip" and "fix it quick and dirty" type of developers. Once you get past a million lines, you can only survive with a very heavy dose of process (documentation, documentation reviews, code inspection, testplan review, automated testing, etc).
The type of organisation that gets this deep into sh*tty code is also the type of organisation that is unlikely to be able to write 10 million lines of good code in 9 months.