 Značim je Daniel Kipper, sem vršan na Raku, sem devaloper v software in sem vzelo v GRUB maintenance team v 2016. V mene je Vladimir, ki je vzelo v počke in izvah v menej vršanih. Vzelo se, da se pripravimo, če smo pripravimo, da se pripravimo, če smo pripravimo, če smo pripravimo, če smo pripravimo, če smo pripravimo, če smo pripravimo, če smo pripravimo, če smo pripravimo, če smo pripravimo, če smo pripravimo. Mi je dobroh vdele, ni je nuance se obragila drzati. Zve dobroh vdele, ni je nekaj takov, neko dobroh vdele, ni jo je tobe dobro. Zve dobroh vdele, ni je nekaj takove, neko dobroh vdele, ni jo je tobe dobro. One was EFI boot service availability for DOSes, and another thing was support for the self-reco table OSM others. These both features are used by Zen to boot it on EFI platforms. And of course plenty of fixes for issues found during development. What are our plans for 2.04? We are going to release that in the second half of this year. And the most important features are related to the security. Currently we are reviewing the core thing, which is needed to these features. It is called Verifier's Framework. It was prepared by Vladimir and this feature is under review. On top of this framework there will be two security features provided. One is the TPM support, which is prepared by Matthew Garrett. It was posted by Matthew Garrett and is under review. And the second feature is prepared by me. This is EFI secure boot support and it is still under review. Second feature is quite interesting for Zen users. It is ZenPVH support. It was posted by Jurgen Gross from SUSE. And there is a lot of work happening around support for ARM 32 and 64 bits. There is also some interest in Spark platform. And this work is mostly done by people from Oracle on the BIAN folks request. It is happening slowly, but it is going forward. And of course there are a lot of fixes. And I would like to ask Vladimir to tell something about new config language. Vladimir, please take the mic. One thing that we learned from feedback is that grab current config language is very powerful. If you want to do something advanced, but at the same time for most users it is very complicated. And over the board for what it wants to achieve and actually stands in the way of many tools. Like the language is cheering complete. So if you want to do any manipulation of config files reliably you would have to solve health problem. And that is obviously not possible. So the idea would be that we would keep the current grab language as a possibility for advanced stuff. But for most common stuff we would like to have something simpler. And there are several goals for it. First goal is that OS and not kernels. Because currently all the config languages present for different bootloaders are always like this. This kernel used with this init.rd and this parameter. Then another kernel, this init.rd, this parameter. And you actually repeat the same stanza over and over for every kernel. And this actually stands in a way that there is no possibility of signing the kernel. Because you have to regenerate it, no possibility of signing the config. Because you have to regenerate it with every update, which is bad. And also it's easy to actually list all the kernels, just directory listing. It's something that we shouldn't have to put in a text file just so it's in the text file. And this leads us to the other goal is signable. It means that perfectly if you should be able to basically just drop the distro, we've already signed the config and this config should work. And another goal is the possibility of efficient multiboot. Because currently the way it's handled, there would be some mass. Either handle it manually with inclusion of the configs or you have some master installation which scans other configs, which leads to a whole set of problems, like updating a kernel you would need to reboot into a master system. It's a mess, so it has to be efficient multiboot. And it should be something easily manipulatable. I'm currently writing a design doc for it, but just to give a small taste of it. Linux OS could be something like a Linux. Then you just tell that disk is equal to this. It means that you want to use exactly the same disk as where this config file is. And boot path equal slash boot. And init rd type equals, let's say. And then perhaps on top of it you want to boot some, you may have some other config figure for some other OS for another Linux, but you don't want to describe this other Linux in your config. You'd rather let it describe yourself. So you add sibling config and you tell disk equal disk type. And you write, let's say, UID equals some UID and equals slash boot slash OS desk dot CF. It would also have some provisions to handle OSes, which are foreign to liberal world. Like, for example, it could be something EFI and you add an entry that disk and then pass. And that would load EFI binary. That's main primary for windows and other OSes that you grab does not support. Obviously, it's not set in stone, but the idea is that it would be something which, so for example, you could up until here, that's something that you could sign. This way you know that you have no malicious argument, no nothing. Also, so I forgot here, argument. And you put any argument that you want to pass to the kernel. Also, the idea is that it would perhaps not be able to handle all the use cases, but for the most advanced you would still have the current config language around. But this is something which should be able to handle most use cases and most common variants. But I especially look for the feedback about what do you think about this idea, what kind of goals would you have, what kind of changes you would like to this idea and so on. And we actually have some discussion time at the end and we can use that time. So, let's continue. What are our more generic plans? First of all, we would like to increase purchase review throughput and decrease response delay for emails. It's quite difficult for us because all of us are involved in other projects. So it is quite difficult to catch up with all emails. But we are trying to fix it and there is a chance that I will be able to be more involved in the reading emails, let's say. The second, quite important task to do for us is to reduce the number of custom patches needed by various distros. To make grab to useable for them without tons of extra patches, which are currently applied. By many distros. This is our goal and if you are able to help us in that project, that will be great. Of course we would like to in general improve overall cooperation with distros and other interested parties. And last but not least, we would like to introduce automatic testing of grab to code from the industry. I don't have any specific idea right now. I thought about something which is quite autonomous as I know Vladimir has some scripts. Maybe we will use that as a starting point, but we are also thinking about something like Travis or something like that. This is not a set in stone, so just idea. And as Vladimir said, we would like to start discussion about the grab project, its cooperation with the community and distros. And we would like to know your opinion in regards to many things. What do you care about? Maybe there are things which are not important for you. And are our plans attractive and interesting for you? Maybe we should change some priorities in the project or maybe there are other things which should take into account and think off. So let's start the discussion, so we are looking for questions. Or suggestions too. Thank you for trying to make the grab reusable. I'm from the City Administrarium Munich. We're making our downstream distribution of Ubuntu. We also have a patch version of grab, because for us it was very hard to, I mean many distributions have some fallback in a drum FS, which basically gives you the chance to drop to a root shell. And we had a hard time securing this for normal client systems. So we basically patched the generation of scripts of the grab conflict. So putting a password on such fallback in a drum FS option is an easy way, it would be great. As usual patchers are welcome, we are happy to review. That's actually the problem that current generation is complicated and almost unmentainable. And also one idea of this is that you don't have to regenerate it every time you have a new kernel. That's why voice and not kernels. So the idea is that it would be something basically readable and human-writeable, not something that only the tools can generate. But I'm not sure that you told that we're also planning to leave the current support for shell language. So there will be both conflict languages available. Yes, I told it, but it's important to remember it. So both conflict languages will be available. And do you have no plans to deprecate the more complicated language at some point or is this still something you want to decide at a later point? It is too early to say. I generally probably know because if you want to deprecate more complicated language you would have to replicate all those features in a simpler language, which would make it complicated. Yeah, exactly. So at this point right now we don't think about that. Thanks. Any more questions? So for me I think the biggest thing that needs to happen is, as you mentioned, reduce the amount of patches that the confusion needs to carry. For me the absolute top one there has to be the irreversible support because that's, it's painful just to watch how different distros create. Yeah, and that was one of the problems that this verified framework tries to solve. I'm aware of what it looks like. Four different verification mechanisms that different folks want for different uses in the whole clash in terms of patches. Yeah, it's painful. And we would like to promise that we are not going to introduce another Linux command to the Linux mess. That would be good too. And I mean, following on to that with reducing the Delta in patches I know it's pain but more frequent releases because that has been in the past. There's some things that have been fixed upstream in time, but distributions don't want to make package something that hasn't been released and hence they keep cherry picking and then, you know, you start with one patch then it's two, then it's 20. There's more than 50 in some distros. Oh, I think Peter Jones actually automated this so he just imports everything. I've discussed that with him and he doesn't like current state so he started posting some patches fixing and so on, so we'll see. We have five minutes more, so... Yeah? I was wondering if you're currently in an efficient state that you can handle all these different requests. Handle what? Sorry, I'll give you... In terms of reviewing the patches and handling the... all the different requests. That says something that we... especially partially we are aware of the promises that we have spoke with other project who had sent us in our problem and we... we actually already got a separate service specifically for whatever automation we don't implement to help us. We would not like to announce it before... before we can have an exact plan but it's in the work center. It's most likely to happen this year. Anything else? So I think that... Could we change the default reply to policy on the mailing list? Because the mailing list works different from any other mailing list. I will try... It just means that people keep getting dropped off. Yeah, I know what you mean. I will try to change that and ask at least my tennis to change that, but it's awful. That's a pretty much flame-boring topic. Because on the other list... I'm so accustomed to this reply to that when I'm on the other list, my replies get lost to the list. So I guess the mailing list would be the place to have discussion, right? Yeah. And right now it has two proposals, discussion and purchase, which is bad. Because, for example, it's hard to... To see which parties are committed and which ones are not. The current state... And also voicing the objections is also not that clear. It may get lost and then it's unclear whether it was addressed or not. There have been cases when we have accidentally committed something where another material had a valid objection. Is there any patchwork help? Sorry? Patchwork. I mean to patchwork. Is there any help? No. OK. I think that we have finished. Thanks a lot for coming. Thank you.