The partitioning of utilities between the ramboot ramdisk and the paqfs means that rerootwin -f ramboot no longer works because rc and supporting files are contained only within the paqfs, and the namespace.save9front file doesnt try to mount this.

The plan9.ini modifications for the livecd seem to be missing from the repo. They perpetuate themselves if the new cd is built from a previous cd install though. It goes to the 'patches vs distro' dilemma. 

The skel.tar file remains a hassle.

When trying to secure the service namespace against failures of the main environment, the mount of cs in the default /lib/namespace file is problematic. Even when using factotum -x, new processes started by cpu and import will try to mount /srv/cs and freeze. The /lib/namespace file used by the early environment is stored in ramdisk so removing the mount of cs (and dns?) will prevent frozen logins via cpu. *Some of this is wrong* It seems that this problem only exists when using auth services, if we are just running with a hostowner key/factotum then remote access works even after loss of the root fileserver. Also, the /lib/namespace file is compiled into the paqfs, it is not just on the ramdisk, so editing it isn't as simple as all that.

the install scripts should autodetect partition locations and edit plan9.ini files accordingly

when controlling linux machines via hubfs, it is necessary to use bash if you wish to use interactive tools like ed. using rc results in input being processed by the rc running underneath ed. why?

------ non-ants specific random questions ------

The u9fs in the bell labs distribution needs two minor patches to build correctly. First is taken from https://code.google.com/p/plan9front/source/diff?spec=svn4be476d6a993298bb5a797a6f872212536c69541&r=4be476d6a993298bb5a797a6f872212536c69541&format=side&path=/sys/src/cmd/unix/u9fs/plan9.h&old_path=/sys/src/cmd/unix/u9fs/plan9.h&old=eaccc3e8d22668badfdfd7fa12c67f3fbb94c67d

in file u9fs/plan9.h in the enum around line 100 specifying 

	Runemax		= 0xFFFF,	/* 16 bit rune */

is one, and the other is adding in u9fs/rune.c in the obvious spot at the top enum

	Bit5 =2;

How exactly does inferno achieve equivalent functionality to plan 9 /srv? chan seems the same but different...

why does it work to communicate with venti on loopback through an import of /net for writes, but some reads are unreliable? (cant fortune)

similarly, why are some binaries non-executable via vacfs? (both seem to be related to short reads, client not requesting more data?)

============= Resolved: ===============

Unless I'm missing something, there is no reason to have separate copies of the mkpaq and toolcopy scripts for the amd64 build. Everything should just be parameterized by $objtype. There's no reason we shouldn't be able to build on other arches than 386/amd64. Fixed in ANTS2.6

there is a namespace/file conflict between /lib of the 9front bootfs and /lib from the ramfs skeleton. trying to run cs in the early 9front environment will complain that /lib/ndb/local does not exist. Fixed in ANTS2.6

the partition and filesystem detection logic in plan9rc should be improved. Fixed in ANTS2.6

cfs support was broken. Fixed in ANTS2.6

The ants bootpaq should probably be integrated with the 9front boot paqfs, with a standard organization. This would be more space efficient, and would solve a few different issues, such as eliminating the need for the one-line customization of rc to use /boot/rcmain rather than /rc/lib/rcmain. This will require careful analysis of the existing boot scripts, which are written to look for a lot of things in /boot. Fixed in ANTS2.8

the -m parameter is not passed to hjfs when plan9rc is the boot script. Fixed in ANTS3.0

root.targz has been removed and placeholder files added for ANTS3.0

hubshell should probably make %remote shells do rfork n to isolate their namespace. Method of creating %remote shells changed in ANTS3.0

There are 3 different copies of the lib/namespace file for the ANTS boot environment in the repository. The one that is actually used is at root/lib/namespace. Why do bootdir.extras/namespace and bootdir.extras/root/lib/namespace exist? Fixed in ANTS3.1

The logic for adding a plan9.ini provided privpassword to factotum is placed in the 'doafterfact' function in plan9rc, with a switch statement which only activates it if an 'afterfact' command is not supplied. There is no reason to make these mutually exclusive options, the doafterfact function should be changed to remove this switch logic. Fixed in ANTS3.1

The venti configuration stored in the arenas is overridden by plan9rc even if the user did not provide parameters. The venti command should only provide parameters if the user did not. The logic in plan9rc should also be improved to not just key off the # symbol but check for /dev also. Fixed in ANTS3.1

The foshalt script should be integrated into a patched version of standard 9front fshalt. Fixed in January 2018 live/install cd release

The build script needs to be decrufted in a serious way. Fixed by restructuring whole repo to ditch labs.

ventiprog should be more careful about not replacing the wrcmd with bad information. Fixed by rewriting ventiprog to be based on backup.example.