mTCP NetDrive "Undo/Go back" feature preview (Can your hard drive do this?)
Ғылым және технология
(Set your video playback to 720 to make the text easier to read)
(Hey Al - better quality and with voice!)
This is a preview of an "undo/go back" feature I'm adding to mTCP NetDrive. Right from DOS you'll be able to set checkpoint marks on a network drive, change things (and possibly mess them up), and then jump backward to a checkpoint mark undoing the intervening changes. The feature is implemented using a journal (log) file that is paired with the original drive image, so the changes and checkpoint marks persist as long as needed across sessions and server restarts.
Most of the magic is on the server side, where a journal for the network drive is maintained. The DOS device driver is unchanged, and the DOS command line utility to manage net drives is slightly larger.
Contrast this to the existing "Session Scoped Writes" feature which also uses a journal, but that journal is scoped to the life of the current session, meaning that when the session ends the journal disappears. That allows for multiple sessions against a single drive image at the same time that users can write to, without having the writes conflict. Think of "Session Scoped Writes" more as a "demo" mode vs. this new feature, which will be useful for real work.
Пікірлер: 12
I have dreams of being able to netboot PCs using a ROM on the NIC and it then loading netdrive and running the system off it.
@mbbrutman
2 ай бұрын
Sorry, requests are at a different department. This is the upcoming features department.
@adriansdigitalbasement
2 ай бұрын
@@mbbrutman hehe!!
Some modern filesystems like ZFS have a similar feature but I certainly wasn't expecting to see it under DOS!
@mbbrutman
2 ай бұрын
It is kind of cheating; the server side is doing the heavy lifting. :) But yes, now we can do it in DOS ...
First Post. Awesome MS-DOS work too!
@mbbrutman
2 ай бұрын
Imagine a Beowulf cluster of these ...
Love it. Will we have the ability to promote a checkpoint to "base" or "origin"? For example, let's say after a few days of working in a particular checkpoint, I just want to commit those changes "permanently". Would that be possible?
@mbbrutman
2 ай бұрын
Not just possible, but already part of the plan. Otherwise you have the problem of the journal file just growing forever, or not being able to use the image elsewhere because the changes are in the journal. I'm also toying with the idea of implementing sparse images, where unused sectors do not need to be stored. However I'll have to implement trim for FAT12 and FAT16 to make that work well.
Very cool. Will checkpoints survive starting and stopping of netdrive on the server, or even server reboots?
@mbbrutman
2 ай бұрын
This new feature is implemented using a per-drive journal file, which is persistent on disk. So the checkpoints and changes survive across your user sessions and server restarts. It's similar to the way cloud storage providers also treat hard drives. The other existing feature *"Session scoped writes" that is already available also uses a journal, but that journal is private to the currently connected session and does not persist past the end of the session. That's more of a "demo mode" feature, allowing multiple people to connect to a drive image and interact with it without corrupting it.
The keyboard needs more buckling-spring. ;)