Originally Posted by
ObserverA3
I've been thinking along the same lines - this is the (in)famous BER approach. Nobody forced them to only take the site down last Thursday (or whenever it was). Having said that, I would have assumed that whenever they took the site down they already had a piece of fully tested software, so it was just a matter of updating the SW, bringing the latest data in and then go live again (slightly oversimplifying but hopefully not much - this is not rocket science).
The usual procedure I saw is that the two versions are running parallel to each other while the inputs are duplicated. In this case it would mean that the data coming from their *A partners and their own flights are put into both systems. After some time you have a look if the results match up - if they don't it's time to hunt bugs.
After that you let some beta testers loose in the system and finally you reset everything to the values from the live system and open it up to everybody. Only this last step should require a system shutdown.
The way the data is delivered has probably remained the same (exept possibly LH), the milage tables have remained the same so I see no reason there that couldn't be tested in advance while both systems were working (the new one on test servers).
The only errors that I can see which you couldn't check for while running the systems in paralell are related to load factors (all the desperate FTLers who want to see everything) or the different calclualtion models for their own flights.
Did I overlook something?
Nevertheless I know that sometimes bugs crop up once everything is live - I don't want to change positions with them ;-)