On Wed, May 23, 2007 6:41 pm, Gergo Szakal wrote: > What we cannot do: > * Make the CGI interface work. > * Make the installer reachable via SSH.
With these two out, the pfi.conf mechanism is the only way to do a headless install, correct? I'm OK with that as long as the pfi.conf mechanism works, though it would be a bummer as I thought the web interface to the installer was neat. (I did the layout, so of course I'd think that) I can understand how it would be hard to match the installer with the web interface, as it depended on an interface abstraction that may not be present in the new version. What keeps the installer from being reachable via SSH? As far as testing goes, it would be neat to do this with a vkernel, but it doesn't sound completely feasible. A virtual environment like vmware/parallels/qemu would do it. Looking at http://www.dragonflybsd.org/cvsweb/src/nrelease/Makefile?rev=1.63 - If you built pkgsrc packages of the new installer, and added it in the ".if defined(WITH_INSTALLER)" section, and changed PKGSRC_BOOTSTRAP_URL earlier in the file to point to the location where they are kept, along with copies of cdrecord and the bootstrap packages, you should have just what you need.
