Right but, then you have to sift through the dozens of unrelated comments, and if there are multiple versions of the patch frequently finding them and cross referencing the changes can be painful.
You need look no further than the versionX->versionX+1 logs some of the kernel developers do to understand the problem. From one patch set to another the conversation around a piece of code gets lost, so much so that for long running patch series people will frequently show up and make the same request someone earlier in the series made because they missed the 20 emails talking about the pros/cons of doing something a certain way 6 months earlier.
There are 7 or 8 independently operated mirrors of the linux kernel mailing list archives. On top of that, thousands of people have local spools of the list that could be used to rebuild it in the event of catastrophic failure.
You need look no further than the versionX->versionX+1 logs some of the kernel developers do to understand the problem. From one patch set to another the conversation around a piece of code gets lost, so much so that for long running patch series people will frequently show up and make the same request someone earlier in the series made because they missed the 20 emails talking about the pros/cons of doing something a certain way 6 months earlier.