On Mar 27, Nathanael Nerode <neroden@fastmail.fm> wrote: > * Most machines have only one interface (If Debian is running on more routers > than workstations, obviously this would be wrong, but I doubt that's the case.) Every laptop with wifi has more than one interface, and so does every system with a firewire port. > * Lots of hardware is crummy and needs to be replaced at least once in a box's > lifetime. Most modern hardware actually has one or two built-in ethernet interfaces, which makes hard to replace them. > found. The proposed solution at the top of thread is pretty good: And as I explained, it is also wrong. The single most annoying problem with udev is all the people who try to fix without understanding how it works. > To be more specific about how this would work, network interface naming would > be a two-stage process, with all "new" interfaces delayed until after all "old" > interfaces were believed to be up. Racy (and slow). > Marco D'Itri wrote "Think harder about it and you will understand why this cannot > be tested in practice," but of course that's bullshit. It's perfectly > implementable and testable -- for testing, the delay and timeout could be set Cool. Send a patch then. Less talk, more code. > quite long. It's not perfectly *reliable*, but it's just as > reliable as anything which depends on udev "finishing" setting up /dev, and > we've been able to handle that (gobs and gobs of stuff depends on that, and > we've managed to live with it). No, we worked around an obsolete model of the boot process by implementing udevsettle, which makes it synchronous again. > Implementation is complicated enough that I wouldn't ask anyone else to do it. > (And it's easier to just work around it than to implement it myself.) You almost fooled me, I sincerly believed for a few seconds that you were ready to support your position with running code. -- ciao, Marco
Attachment:
signature.asc
Description: Digital signature