[tobiasu]

This discussion comes up every time only because some people seem to think OS development is like racking new x86 servers running RHEL.

Many of the machines do not have LOM. They have hardware failures instead. They hang because they get trashed building OpenBSD and ports pretty much 24/7. There is debugging and serial cables going on. Someone needs to push that NMI button and check the LEDs flicker like they should. Reboot them. Constantly update to the latest development version, making them panic quite a bit. Diagnose that. Installation procedure requires console access, monitor adapters, weird keyboards, ... They don't fit in racks properly. There are security concerns. Etc, etc.

It's wrong to think of the machine room as rack space than can be had for cheap somewhere else. It's much more like a lab (with the mad professor living on top, controlling the experiment).

[myrandomcomment]

While what you say is correct, Theo's stance on this is still a bit unreasonable. A review should be done to see which systems can be moved or supported by the means of remote power off strips and IP console servers. They should be perfectly willing to move that gear if someone offers them the space. All the Sun SPARC, Alpha and Intel most likely falls into this category. Only systems that someone needs to be physically there to access should be left onsite.

I have donated to OpenBSD a number of times because I believe the project is of great value. In all cases where I used a release (for firewalls mostly) I purchased a CD set.

[Istof]

maybe he doesn't want anyone else to have physical access (for security.