 So, okay, I am reducing and letting talk that I actually wanted to do so well at the BUNTA conference and it's about making my job super-fishes and First I have to warn you, I have no idea what I'm talking about and what I'm doing. So I Don't know for example much about butter affairs and stuff like that, but I'm a big fan of git so I try to use it or have some thoughts and A proof of concept on how to use it for packaging so The old way to do everything is everything is a file so when you install something you Put a lot of files on a system and then you have a new thing there, but well It turned out that everything was changing rather quickly because not everyone is seeing the same file File system tree, which is for example with things like Docker and stuff You can actually have things that only some some parts of your system see the other thing is we have git which Is awesome and is versioning and Is it content addressable file system? The cool thing about that is that every object in the file system actually has has the hash and that identifies it uniquely You can abuse it quite easily. We did that with by the sector the gnomes guys did that with OS tree where they put essentially the whole operating system in a binary repository and Yeah, actually one might consider doing that for packages to to have delivered packages like that and One reason for that is we'll get you have atomic updates because right now packaging looks like that You put the batteries out and you fall on the new package And you really hope that everything goes fine, and then looks like these get on the right side This is how you hope it works, but sometimes stuff like that happens and Then you really want to really want to have The ability to go back atomically and everything like you had it before so The other thing that's the one thing you want with atomic updates the other thing is this and All the web applications are doing that so everyone Has that in the web world so that you just up don't stuff on the fly But I wondered why should that really be restricted to to web applications? Maybe you can do that on the desktop too so I did a proof-of-concept and it's really slow, but it works and What I did is I created a few surface thing in Python, so yeah, it's slow and what it does is it checks out Or you can run you can mount a biosecond repository so you can mount it get repository with LibreOffice builds and Then you can say okay. I want to have this built and you can start it from there and what happens is that Whenever you open a file it actually while when only when you are opening it It creates the file locally in a cache and that gives you a handle to that so with that you can Essentially have on-the-fly updates in the background because you can push new things into this repository Without ever changing the file system stage In the first I also try to have two two bills doing the same thing as to To LibreOffice is running on the same file system But that really doesn't work because the dynamic linker really gets confused when there are two different libraries with the same file name around So it should do a little bit of So I skip the demo because I have no time left, but Yeah, it's only like a three-on-one Python thing So yeah, that's that's how I would love Us to maybe do packaging at some point in time and not in the old way falling backwards on the backwards Okay. Thank you