 Tako, to temo je na bojdu, ker je nekaj, nekaj, in nekaj nekaj, za intrepatičnjih vzivov. Bila biti na vse tudi, da nekaj, nekaj, nekaj, nekaj. To je vse, ko ta, ko je, ki se nekaj, nekaj nekaj, nekaj, nekaj, nekaj, nekaj. zgod Maybe the ex linus would be better. So I tested some botloaders and when we get into discussion to see what are the problems, then what are the solutions tako, da bomo poživati, da bomo načinati, da je reličnji reličnji, in da bomo poživati, da bomo poživati Lilo. Na početku sem pripravil nekaj, nekaj pripravil nekaj, nekaj nekaj, da sem pripravil nekaj, zato je bočnjenek ko je jeziojeva na vsezvanje vsezvanje, vsezvanje, kako vsezvanje, vsezvanje, vsezvanje, vsezvanje. In začin, kako smo vsezvanje, imamo armarje in vsezvanje. Ne bo se, da se zelo, da se predaj pričal o to, In vseč je, da bioče počkali cpu v velim mordu, v 16 biti, in tudi na 20 ljudi, da je inoče počkali, da je počkala, da je nekaj kompatibilitiv na primeru procesu intel. 8086, 8080. V sezlušenju, ne boš, nekaj zelo v 10, 15 rov, boš nekaj, boš pa vsezlušenju, in je vsezlušenje, kaj je zelo vzelo, Tako je zelo, da je tudi zelo, da je osega in velika interfejsa. To je zelo, da je stavila. In potem je osega boota. Now the modern keyboard, the modern motherboard that try at the beginning to boot with the network with PXE, both P-protocols. And then they check avalit media that at the end of the first sector, this magic number, 55AE, the BIOS will load this sector, so it's important to know that we have only 500 bytes to load the rest of all. The operating system, so they are not very much seeing that we can do about different hardware, different file systems. The old method that was in the PCDOS, for example, there was two type of boot sector, the master boot record like. In this master boot record we have four partitions, and the loader try to find the active partition and boot this, load the boot sector, the first sector of this partition and boot this partition. And now, since a lot of years, we can have a lot of more partition, and this partition are partition inside over partition, so in case of five or more partition we recourse this method. Or there are the boot sector, in those there was some set up of BIOS because of floppy parameters, and then it would try to load to find the EOS system or MSDOS system, it depends on the version. It will try to locate this file and to load this file, and then it will execute this file, and this is very similar to what we will, what extra Linux will do. It seems that it try to simulate this old DOS behavior. Then we have also Linux, we should boot Linux. Now the, yes, we still have the boot sector at the beginning of the Linux image. We can test it with, and we see the signature at the end of the first sector, but as far I know it's not usable since the, I think the version 2.0 of Linux kernel, cannot boot this way, and now it's also too big, and yes it's, I don't know if somebody tested this method, and the old method was to use RDF and similar tools to set up the parameter, the root partition, direct in the kernel image. Then we had Linux loader, the first Linux loader for Lilo, and it has the advantage that we can choose what kernel to boot, we can set up the parameters on the boot prompt, and we can boot also over a operating system, and it works also for the hard disk, so it was a nice improvement, but still with a lot of problems that we see later. Then, yes, the boot loader will load the Linux kernel, and the initial root system will execute these, kernel will, first it will decompress, it will set up the CPU, it will load the loader, it will execute the first, it will mount the root system and execute the init, and now we had the initram file system that it's executed by this init, it will tout lot the real root, it will then change the root and continue the booting process, so the art. Already a lot of step and different program, different maintainer, different view of how to boot Linux that might be in some time later should be unified. Then I, okay, so we go to the real part with the boot loaders, and here I want to have some feedbacks, some discussion about problem, because the main problem is that there are so many different type of use with server and desktop that it's very difficult to know if there are one single boot loader that can do all things. I forgot to tell you that now I am speaking only of the PC architecture, so the E3, H6, and the, no, AMD64 architecture, at the boot time they are the same architecture, and totally often, yes, on the PC architecture not on the Mac PC that use an overabjose and overboot sequence, and not on the over architecture. Also not, it would be nice to have also one single boot loader for all the architecture, but I don't think it's possible to have a single interface. So they are Lilo, it's the first boot loader. The last version come in February 27, 2007, and now they are no upstream website available. They are, it seems that, yes, the upstream is dead, but Lilo has this problem a long year ago, and sometime we managed to, we, the Linux community managed to find some new maintainer. It seems that now they are no, we are lack of maintainer, and this is the big problem, because Lilo should have some additional development and support, for example, for the multi boot protocol. Then we have a group one, the older group. It's not only for Linux, it work also for hard and a lot, and it work for a lot of architecture. Also in this case, the real last upstream version is very old, 2005, but there are still some activity in subversion, and they can take new patch, but upstream don't want to include new feature in this group loader. So they are the second generation boot loader, but it's not so used. Also the last official version is very, it's not very old, it's old, it's from 2008 in GPL3, but there are a lot of activity in subversion. Now there are a lot of activity. Also in this case, it seems that it's not so sexy, seems to have to work on a little boot loader, so we see that on some years they are fast, they are fast not development. And then there are the last born boot loader, they are the CIS Linux. It was already used, since a lot of time to boot the CD-room for network loading, but now we have also the X-Linux, that it seems that it try to simulate the old dose behavior, so it's very small boot loader, but with few feature, but it's the nice things that it was used for the CD boot loader, so it has a lot of support for help, for user interface before loading the real operating system. And also the nice feature is that it's developed by H. Peter Harvin, that it's the maintainer of the kernel boot sequence. And it's no very well BIOS and interactions are spent to disk on Linux kernel, so in this case the boot loader and the kernel development are always compatible, but only for Linux. So the design, Lilo used a very simple design. At runtime it will find every sector of the kernel, the number of the sector, the BIOS encoding of such sector, and it put in a file. So the Lilo had a very simple task to load the sector, already encoded in the BIOS schema. But the main problem is that if one overwrite some kernel, some kernel, then the Lilo cannot boot anymore. And this is very annoying. Now then there was some development on Lilo. We have a menu, we have a password protector boot, but it's not so feature, we asked the group one and group two. Group one and group two take a different design. They have maybe a big code, in this case there was a big code in one or two stage with lot of the grab loader. But when we have the group, we have the prompt and it can read a lot of partition, it can read the kernel some file that was not available when we install group. So in this case it seems that it's difficult to have an on working booting. But as we see in the million list there are still some problems sometimes. And last is the X linux. In this case it's very small, it is specially set only to read X2, X3 partition and it only so the support is very small, it takes not a lot of bytes. So for this it's a very nice boot loader but at very few feature. And it works nearly automatically, so we don't need to run X linux every time we install a new kernel. But in some case with linux logical volume management and the rate and some rate configuration where we should be sure that there was in the right mirror there was the kernel, there are the kernel. But these are special requirements. So now the thing that we could start the discussion. I think that boot loader should allow us to choose the kernel to provide the kernel parameter when something goes wrong, when we forgot the root password or yes for new kernel. Sometime with X we should sometime debian stable is unstable so we should have the single mode. And it should allow the initrama, loading the initrama file system and so also big images. I think a nice feature was the kernel, we should be able to run the kernel, a new kernel only once and then go to the default. This was the case of Lilo and it took a lot of time to group to have this feature but now they are group as this feature so it's not mere blocker. Should be usable by system administration error. This not the real professional system administration that I hope they will not make such dummy mistake but for the normal user that will need to test new kernel. Then it should be easy to install and this is my opinion, install and run the memtest 86 because a lot of time we saw memory bugs and we give, yes we file bugs about kernel, about X when there are no real bugs. It should be easy to configure and I think now I think that the actual loader are not so simple, not easy to configure. I think we should do something better. There are many people who are trying to get along and do something, although Lilo does have some disadvantages, at the very least it's properly documented and the document describes what the program does in excellent detail. Yes, I agree that documentation is important, then maybe it's also a problem of the community because I should say that a lot of time I don't use the official Lilo documentation but there are a lot of nice auto and over documentation. But yes, for grab 2 this was a big problem and it seems that now there are a lot more pages but I didn't control accurately because the problem of the grab group website is it has a lot of links but then they don't work. And yes, it's a problem I found yesterday that now group 2 had a nice documentation on what to do when things go wrong and it was lacking, it was a big problem of group 2. And X Linux has a lot of documentation on how to write new interface, so for developers, but it lack also in this case the documentation for user and how to use for the normal desktop case. Ok, so the Lilo, the problem is unmounting, it's don't handle the big kernel, it don't support the multi-bot protocol standard that it's required back then and yes, some people forget to run Lilo. And it seems that Manoj remove the post hook script in the kernel package because it was so buggy and so we should find also some automatically tools so that when we install adebian kernel Lilo will automatically be run. But I should still discuss with Manoj because yes, it was a very recent change I think and group 1 is no more really maintained so we cannot go to group 1. Group 2, the problem is the training of whole system in, they are not as Iana said, there are not a lot of documentation. It use non-linux partition notation so it's also a problem that a lot of time. Personally I am off one, I have the off one problem because all partition disk start with zero and we are, I'm using always the first partition is one. But okay, it's my problem. It's not add zero comma zero for the first partition in group. Yes, yes, and what is the zero partition? It's the, yes, yes, yes, yes. Partition schema notation, ada1, but it has one notation. Yes, yes, but people will get starting to use. Yes, and I'm still wondering if group 2 will be released. Yes, I don't know if you know some information. I don't know because they are working a lot on it. There are a lot of working on it. I can make quite an announcement because yesterday or two years ago I talked with group 2, I received and I said hey, if you don't code password support it won't be used by mainstream distros. By mainstream distros. So they have begun yesterday or so to code password support. So it's, you know, if you don't have password support in a school even if group 2, even if the distro has group 2, the people, the seasoned people would switch to group 1. Yes, yes, for sure. I don't know when it will be released. I hope that it is released when it is ready. Yes, but some, it would also nice to have some new pre-release version in order to know that starting from 0.99 it has such support, 98 has such support. And the big advantage of group 2 is that it supports a lot of architecture, not only the PC. It supports a lot of file system. So it's, yes. I don't like it, but it is great. And it supports, natively, the over operating system. Over kernel, so also the free BSD kernel and such. And then the last is X Linux. Yes, the two is, I don't know, sir. It's very file system specific to have the assembly code very small. But then there are a lot of different program for different file system. It's too new so that there are not so much documentation. We don't know exactly where they are the problem. I don't know if it could work on the server solution on schools. And it supports also in the name only Linux. Really it has the multiboot support so it can boot over operating system. But, yes, you cannot install by now X Linux from over operating system. So, yes, it can boot over operating system but it's not really, it's only a Linux only system. The advantage, we already use it for the CD. The maintainer, it's know a lot of this boot sequence. So, then the question is we should remove Lilo. We should have only one supported bootloader. And at the end you have some doubts if the lack of multiboot support is a problem of Lilo or a problem of Xen. Because also in Xen fact they say that there are lots of people and they don't have time to implement, to really implement Linux image to boot. Maybe it's also not so difficult to create this image because we have only to add small support about self-encryption, self-decompression in the initial sector. So, now you should tell me if you want only one bootloader, if we should remove Lilo. Is there any feature that Lilo has that Grab1 doesn't have? That's the first question I can think of. Grab1, I think it don't have the feature to load a kernel only ons. And I use a lot of feature for remote server. When I should test a new kernel, I put in the parameter panic5. So, if it cannot boot, then in the next boot I will really work in kernel. I think Grab1 don't have it, but Grab2 has this feature. Well, in fact, Grab2 has this same feature. It might not be documented, but you can do it. In Grab2, yes, but also Group1. Yes, you say that default and fallback. Ah, yes, and then do over. Usually, in order to set default entry, you use a number. So that zero is the first entry, one is the second entry. If you use save it string, it looks into save default file, which is found at slash boot slash group. And so, when you boot a kernel entry with panic equals 5, you can choose before boot in this kernel to say save default is 2. And 2 is another boot entry. So, the next time you reboot, it boots entry 2, because you are saying default saved. Yes, yes, it can be done. The only thing is that it has been reported several times that this does not work. So, I suppose it is a book. It does happen in some setups, in some distribution, it happens more frequently. So it does work, but if you are lucky. I mean, because I have read a lot of forum support questions about this issue, and sometimes it does not work. I do not know what is the Debian specific status. But to answer, I think that in any case, we should not go to group 1. If we should choose a boot loader group 2 when it is ready, but if we will now go to support only one boot loader in a few years when we should change again, I do not think it is so good. Well, that depends on the freeze. If freeze is going to be soon, grab 1 is the one that is working. Well, I think that using grab 2 is a good release goal, but if the freeze is happening too soon, that won't be able to... Well, we have the proposal, but I'm not that sure. Yeah, yeah. It's on the boot page. Oh, okay. I won't start a flame there. So group 1 is already the default boot loader. So there's no other chance, I think, I'm afraid. What? There's no other chance, I'm afraid, because it looks that group 2 is not ready yet. Yes, but yes, this discussion is not for the next release, because also I don't think that we could remove Lilo for the next release. But it's more for the future what we should do. And we will... But I don't think that we will solve all the group 2 problem. And I don't think it's so important to... We still have... Now we have a working boot loader, so we can continue to have group 1 for the future. I wanted to ask the question about... Is there any setup when Lilo is needed, because group 1 does not do the job, the work? So, yes. There are some strange LVM plus write setups, or whatever it is, maybe a new file system that is not supported by group yet. In this case. And what you can do with Lilo is just install it, drags down the section number, and whatever file system, whatever strange thing you have, it usually boots the kernel, because it just looks for the specific sector and boots it. So it's quite boot at it. I do not like Lilo at all, but in some setups Lilo does work, and group 1 does not work. Group 2, it depends on... Group 2 works better, but it depends on exactly the setup. That's my opinion about it. I don't know if somebody of you know exactly what setup work on Lilo and not on Group, because every time we discuss the Lilo removal, there was some case in the mailing list, but we don't know exactly what was the problem. Could you pass the microphone? So, one thing that's quite obvious that works with Lilo and doesn't work with Grub, is if everything is on LVM, and the reason for this is that Lilo encodes the actual block numbers of the pieces of the bootloader, and it asks the Linux LVM system at sort of map install time. So that means that the bootloader itself doesn't have to understand LVM, whereas I think, if I'm not mistaken, that even Grub 2 still does not understand LVM 2 metadata, because it would need a parser for if it were to support slash boot on LVM. In fact, Grub 2 does support some kind of LVM boot. I'm not sure about it, but they are working both in LVM and in write support, and have LVM plus write support in final Grub 2. The thing is, I do not know what is the current status of what it is actually supported and what it is not supported. But there is some kind of LVM support in Grub 2. But you have to check in SVN in the mailing list and ask them the lack of documentational groups. I want to work on that. So there is a utility called M-boot pack, which can be used to take a set of multi-boot images and turn them into a thing that is acceptable to Lilo. So I've successfully used that boot Zen with Lilo in a configuration where I wasn't able to get Grub 2 to work at all, because I don't know exactly what was wrong. But with default Debian Zen kernel, because I've seen that the problem is... I've worked for Zen upstream, so in this particular case it was an upstream Zen. Yeah, it was a perfectly ordinary Zen multi-boot image in a Linux multi-boot kernel and in-it RD. And you run M-boot pack and you get a single file that's got a little piece of code at the front that sets up all the multi-boot stuff. Because Lilo had also the problem of very big image. So when we have big kernel, big in-it RAMs... Well, evidently, I didn't hit that particular problem. But I see that it was corrected at the beginning of this year. I don't know, but I just thought I'd mention that. It has various advantages as well because it seems somehow wrong to make a copy of the kernel image. But actually what you do is you make a copy of the kernel image that then is owned by the boot system and then when the kernel package is updated, your system doesn't become unbootable because the old M-boot packed file with the old in-it RAM FS and the old kernel and everything is still there. M-boot pack, all one word, stands for multi-boot pack. So we should add something also in the Xen documentation, Debian Xen documentation too. Yeah, that might be worthwhile. So some other comments. There is another... Well, this is a personal opinion. It's about the CD-ROM Debian installer. It's about CD-ROM Debian installation... Debian installation CD-ROM. It currently boots with isolinox. So I think that in two years' time or whatever it is, we should switch to Group2 so that we have Group2 in all the ways of booting the distribution, even in network booked because it is going to be improved to the network stuck in Group2. And the only drawback that I see in Group2 as the CD-ROM default boot loader is blind people get blind people to write boot options because in isolinox you get prompt and you can write as fast as the CD-boots. But in Group2 you have to do some kind of type in C letter and then if you have to boot by comments you have to type to write a lot of things. However, I think that as long as Group2 has script support and a script might be written so that it is the same thing as isolinox so that you write installer so that you want to install and not let's say you don't write rescue because you want to rescue installer, space and whatever options you want. So there should be a way in Group2 to allow direct typing instead of having to type C letter to write the boot options so that blind people can use it easier. But it would be a huge task I think to have group to boot the CD because as I read in the isolinox mailing list there are so many KRAP BIOS CD ROM driver that you should wait, there are so many special keys that to make all people happy it's difficult. But it's nice to have this option and maybe to move. The thing about KRAP BIOS is interesting because Group1 also have CD ROM boot capabilities and I think that it is not Group1 not having that nice graphical boot as it has Korean isolinox boot. I think that the main problem is that it doesn't... But now booting from CD or booting from CD image? No, booting... Directly from CD? Booting from CD, not booting CD ROM is image. No, I'm not talking about that. I'm talking about why Group1 is not at CD ROM boot? The answer is because of KRAP BIOS and Group1 maintainers didn't want to deal with these KRAP BIOS and they wanted to keep the code clean so they don't try these KRAP BIOS codes but they don't work around and so on. Group2, I suppose that we'll have better CD ROM boot support but I do not know what's the current state regarding these KRAP BIOS. I don't know what's the policy. If you have no questions, I'm going to talk about another subject. It's about boot images. We have five minutes, but you can... Group2 can boot... Group2 can read inside ISO images so that you can have USB pendrive with several distribution issues and... Well, you may say, you can boot the ISO image. No, you can... Group can look inside the ISO images, can track the kernel and the init RD and if the kernel and init RD has ISO 960 loop support, it can look inside the CD ROM ISO file 2 and boot it. So it is quite nice feature that Group2 has and I think that no one else has. There are some distros that support booting from ISO loop back, but there are not too many right now. So we enter the discussion and I will try to summarize some of this discussion for the main list. OK.