[lnyng]

> Not regress install time.

They seems to achieve size reduction by transmitting only the forward upgrade patches, and let the machine generate the downgrade patches during the upgrade. How do they manage to keep install time the same while this definitely uses more resources (cpu, io)?

[arcticbull]

Maybe they generate the downgrade patches as the upgrade patches arrive - but before they're applied? During that period one would expect the network to be the bottleneck in all but the most extreme circumstances. [edit] And it should be possible to do incrementally.

[rconti]

I wonder if perhaps "install time" contains "download time", and by taking some average internet connection speed, they're estimating that they save more in download time than they cost in actual install+downgrade patch time?

This would mean, if you have a fast link, the experience is somewhat worse, but if you have a slow link, the experience is better.

Just a total guess though.

[reed1234]

My guess is to install an update windows has to apply each little patch to random files scattered across the OS necessitating a file read and write for each little change. But generating a reverse delta should only take a transform of the update file which I imagine is much less complex.

[forgotpwd16]

>How do they manage to keep install time the same

Do they? Or Windows updates are going to get even slower?