> Note that this algorithm also mimics almost exactly what Sun
> workstations have done since almost the beginning too.

True, but you should also consider the general user culture differences
between Suns & PC's. On PC's it is expected that keyboards can be plugged
and unplugged at almost any time, and the system should cope. (Never mind
_why_ people found it necessary to do that often enough for it to be an
official feature...)

Just last night I benefited from this: I was installing a new card into
an ATX system (ironically, not an i386, it was arm32) and forgot to plug
the keyboard back in; I'd removed it while moving the system around to
get the case off. Because of the standard boot block behavior (probably
cribbed from i386 anyway) I was able to plug in the keyboard and go.

It was rather nice to realize (after it booted) that I was able to do that.
Resetting the machine to get it to recognize the keyboard would have been,
well, distasteful.

The default choice for bootblocks in sysinst should match whatever the
existing user culture is for that hardware. If I deliberately change the
bootblocks, that's my business.

> I've been handing out copies of my boot blocks to many people in the
> past year or so and everyone has been extremely happy with them.

Realize that those people are already self-selected by the fact that they
were unhappy with the standard boot blocks. They should not be assumed to
represent the general user base.

You've shown that your boot blocks are good for people with setups like
yours, but you haven't shown that they're good for everyone using i386.
And I don't think you'll be able to, because the wide range of environments
happens to present mutually conflicting requirements to the boot blocks.

Todd Whitesel
toddpw @ best.com