rid of most of the places we use referrer. Also get rid of login modal,
use regular login page, which still allows a referrer. Add a regex to
check the referrer; since the referrer always be one of our pages, we
can use a narrow regex to make sure it is clean.

work!

Apparently, for the last decade or so I missed the part in Linux mount about:
     "Note  that,  depending  on the filesystem type, state and kernel
     behavior, the system may still write to the device. For example,
     Ext3 or ext4 will replay its journal if the filesystem is dirty.
     To prevent this kind of write access, you may want to mount ext3
     or  ext4  filesystem  with  "ro,noload" mount options or set the
     block device to read-only mode, see command blockdev(8)."

resources into different graphs.

into their own graphs. But then a wildfire came along, and it was too
smokey to go outside, and I got to thinking, lets get rid of mouse wheel
zoom cause its basically unusable, and instead add some zoom in/out
buttons. Good thing the wildfire got bigger and the smoke worse, cause
in reality, no one just decides to change anything D3 without an extra
smoke day.

D3 changes are in commit c785048c

Mouse wheel zoom is incredibly annoying when a graph is embedded in
a larger page, and you routinely use the wheel to scroll. You hit the
graph and the scroll stops and graphs starts zooming in and out. Been
driving me crazy for a long time.

A professional D3 programmer would probably have done this in two
(long) lines of D3. I am not a professional D3 programmer, I barely
rank as beginner, so I needed a lot more lines.

user/project show pages.

That clever update status on a single line using \r? Doesn't work
so hot in the linktest window where every update is its own line.

for the case that a reservable node is reserved, but is asked for in the
rspec by type.

Whether to use compression or not is now dependent on the setting
at the top level dataset of the pool. All zvols created within will
inherit that setting. This allows us to experiment with compressing
the underlying volume for a dataset.

This brings management of blockstores to the same level as management
of nodes, interfaces, and wires; i.e., one level above raw SQL queries.

Handles adding or removing a hard drive, including the blockstore
related tables and the hacky "?+disk_foo" node features.

Wanted to get this done before I once again forgot what needs to go
into the DB to make local blockstores work on a node.