WEBVTT 00:00.000 --> 00:11.000 Thank you, thank you. Thanks for the introduction and it's a great pleasure and great honor 00:11.000 --> 00:18.880 to be here again. Yes, this is the last talk of this in the room so it won't be too technical 00:18.880 --> 00:25.200 so don't worry if you are already tired and why talk about this, why trying to do this 00:25.200 --> 00:33.120 perspective? Well, it's really the 15th of the room this year so 15 years depending on 00:33.120 --> 00:42.400 how you counted since Jakubirmar started this tradition in 2012 and there have been some 00:42.400 --> 00:54.160 166 talks by 80 distinct speakers about 45 distinct projects during this the room in 00:54.240 --> 01:02.480 the past 15 years including older unicorns and other friends that we have and 15 years that's 01:02.480 --> 01:10.320 quite a lot of history already I think you know 15 years if your child was this stall now it's 01:10.320 --> 01:17.360 probably graduating from Guy High School right so maybe we should look back well even the world order 01:17.360 --> 01:27.840 can change in such a time so what probably didn't change in all those 15 years I think what 01:27.840 --> 01:35.280 didn't change is the significance of micro kernels and I think that this this audience is 01:35.280 --> 01:42.000 approved to that that we still believe in this idea now different people have different 01:42.960 --> 01:48.560 different reasons to work on micro kernels therefore we have different micro kernels 01:48.560 --> 01:53.680 emerging systems and that's perfectly fine but I would say that if there is one single 01:53.680 --> 02:00.480 underlying common denominator it's the fact that we believe that from this software architecture 02:00.480 --> 02:07.200 the micro kernel approach is just a little bit better than the monolithic approach of course 02:07.200 --> 02:11.760 there's nothing wrong with having lot of features in an operating system but the problem 02:11.760 --> 02:19.680 atic part is when all those features form a single point of failure and I think Norman Fesca 02:19.680 --> 02:28.560 really describe it very nicely 15 years ago using his bull gun shift or castle ship analogy 02:29.920 --> 02:40.640 which I have redrawn using AI like this so you know monolithic systems looks like this beautiful nice 02:41.120 --> 02:48.400 a lot of diversity a lot of features but everything is on this single wooden ship without any 02:48.400 --> 02:54.800 bull to tight bulkheads you know do you really trust this is this really the best way we should 02:54.800 --> 03:02.400 implement our software a different point or different take on this if you if you look on the 03:02.880 --> 03:10.320 size of the Linux kernel binary plus some of the required kernel modules not all of them just 03:10.320 --> 03:17.280 those that are required or not typical the machine you get to something like 150 megabytes of code 03:17.280 --> 03:25.440 and data that's comparable to other monolithic software packages like triple aid games or software 03:25.440 --> 03:34.800 sweets and things like that all right but would you run that in a kernel mode would you allow 03:34.800 --> 03:41.040 would you trust your web browser to directly access your hardware without any any you know 03:41.040 --> 03:50.880 moderation layer I leave the conclusions too so what did change in those 15 years and this is 03:50.880 --> 03:56.560 quite important I think and maybe you might say that I'm overly positive but I really believe this 03:59.760 --> 04:09.120 since 2012 micro kernels are actually in the mainstream now you have multiple real-world operating 04:09.120 --> 04:15.920 systems based on the micro kernel design that are running on millions of smart devices you have 04:16.000 --> 04:24.720 millions of cars running micro kernels as hypervisors you have millions of smart appliances running 04:24.720 --> 04:31.280 micro kernels as the user facing operating system like google fuchsia you have millions of 04:31.280 --> 04:40.640 smartphones a lot of running harmony always next so I think I hope I believe that people are finally 04:41.600 --> 04:50.720 you know forgetting all those unfounded anxieties about micro kernels I mean it's not always ideal 04:50.720 --> 04:57.840 but I think there's some some progress and instead of that they are embracing their true potential 04:57.840 --> 05:05.680 without without worries so we have seen some wonderful demos today already I have just a couple 05:06.560 --> 05:18.000 of snapshots here so this is google nest with google fuchsia this is harmony always next running on 05:19.520 --> 05:28.240 surprisingly not so not so recent but but still very nice Huawei phone so yes this is an actual 05:28.240 --> 05:37.360 micro kernel based operating system with all the user facing goodies you would expect and 05:37.360 --> 05:44.720 actually the framework the RQI if I'm not mistaken is the same that that has been previously used 05:44.720 --> 05:56.240 by the Linux based harmony OS so so you have even that continuity so to sum up micro kernels are in 05:56.240 --> 06:06.560 the mainstream now let's take a short trip down the memory line and let's look on some of the 06:06.560 --> 06:12.720 interesting talks that have been presented here at the micro kernel the room of course this is my 06:12.720 --> 06:21.760 selection so if your favorite talk is not being mentioned that's not personal that's just just my 06:22.160 --> 06:30.960 selection but I really recommend you if you have the time going to that old talks and 06:30.960 --> 06:38.320 you might think you might find some interesting ideas there still so this is my this is my selection 06:39.360 --> 06:46.480 in the first year most there were mostly introductory talks so if you would like to compare the 06:46.560 --> 06:56.400 basic architecture of g note Nova Helena was feel free to go there g note already presented 06:56.960 --> 07:04.400 that there is this agony of choice in case of g note because it's not just an operating system it's 07:04.400 --> 07:10.800 an operating system framework that that actual supports I don't know how many it is today but 07:10.880 --> 07:18.560 six seven micro kernels and even a monolithic kernel which is interesting I mean the the type 07:18.560 --> 07:24.320 this is coming from from the original title I don't think it's wrong it I think it's great and I 07:24.320 --> 07:30.400 still remember this quote from the panel discussion by Jacob Germas where he said that Helena was 07:30.960 --> 07:39.600 implements the MFS driver so that there is a migration path for the minics three 07:40.560 --> 07:47.520 Helena was I think it was meant half jokingly back then but yeah I think it's less of a joke nowadays 07:48.560 --> 07:57.840 so this is this is the photo from that panel session 20 13 we have learned that Gidoo Hort is 07:57.840 --> 08:04.640 alive and well we have we have seen a confirmation today but at least for me that that was a surprise 08:04.720 --> 08:12.800 that this operating system based on the on the first generation micro kernel is still quite 08:12.800 --> 08:20.800 quite working and progressing they have been several talks on learning from past mistakes which is 08:20.800 --> 08:30.960 always good and there were several talks describing how operating system projects are 08:31.040 --> 08:41.040 maturing from academic endeavors to practical usability and maturity and probably the most important 08:41.040 --> 08:46.960 point from those talks is try to eat your own dog food whenever possible it really helps 08:50.400 --> 09:00.800 2014 this was the year when Anticanta announced ramp kernels it was them which I believe made a 09:00.800 --> 09:08.160 huge impact Gino would also introduce they their base HW their own micro kernel 09:09.200 --> 09:17.680 Helena was got a sound stack and L4E got commercial so again another move from academia to 09:19.280 --> 09:27.120 to to the real world so to speak. Jacob Germas announced my paternal dot info which is still an 09:27.120 --> 09:34.720 interesting site hopefully that that sort of points you to all the existing open source micro kernels 09:34.720 --> 09:40.720 I'm currently trying to maintain it if there's your favorite project missing please let me know 09:40.720 --> 09:48.640 sense and the poor quest and there was a very interesting talk that I really believe we should 09:48.640 --> 09:57.440 review from time to time by Vasily where he made a very good case for implementing practical security 09:57.440 --> 10:06.080 measures even in micro kernels things like not executable memory address space randomization 10:06.080 --> 10:11.920 and stuff like that because the world is complicated having a nice architecture that sort of 10:12.800 --> 10:20.720 prevents certain types of issues by itself is nice but there are other types of issues that 10:20.720 --> 10:33.200 should be also prevented I mean just recently L4E was in the middle of security vulnerable to that 10:33.200 --> 10:41.440 was not caused by L4E or the micro kernel but the vulnerability was caused by improper use 10:41.440 --> 10:51.760 of that building blocks and yeah maybe if there would be even more strict security measures 10:52.400 --> 10:59.120 maybe those people would simply not be not be able to use it in such a wrong way I don't know 10:59.520 --> 11:13.200 2015 SEL4 joins our community which was nice there was also there were also two talks about 11:13.200 --> 11:20.320 let's say small projects maybe a single person operating system projects but it's always 11:20.400 --> 11:32.880 great to see people learning by doing and while doing that discovering new ideas that we 11:32.880 --> 11:40.720 old-farts just have forgotten to discover or just fail to discover and I think this year's the 11:40.800 --> 11:50.000 room follows that tradition nicely and actually Al4E implements stake galleries I'm not sure if 11:50.000 --> 11:59.200 this was a direct direct direction to Vasily or not but yeah we've we've done something to improve 12:00.320 --> 12:04.480 improve the security even without changing the architecture 12:04.560 --> 12:13.680 2016 G note announced there's support for SEL4 so extending their set of supported 12:14.240 --> 12:22.400 supported micro kernels by yet another one good newsboard gains a lot of usability features 12:22.400 --> 12:30.240 that we still see the results today thanks to ramp kernels like sound support and USB support 12:31.120 --> 12:40.720 and Gianluka also introduced his own micro kernel again presenting some fresh interesting ideas 12:40.720 --> 12:47.280 on scheduling and what you can even consider to be a schedulable entity in an operating system 12:47.280 --> 12:58.400 kernel 2017 probably the largest attendance of a single single talk was happening here when there was 12:58.480 --> 13:05.120 the first introduction of the redox OS to this devroom and this really shows that it's never 13:05.120 --> 13:10.640 right to come to a party I mean you can always come up with something new something interesting 13:12.480 --> 13:17.360 I don't know why surprise I'll be making more popular than Linux 3 at least measuring by 13:17.360 --> 13:24.320 the number of downloads so you see the the migration path for the Linux file system is really needed 13:25.280 --> 13:33.040 and the new world talk showed a nice way how you can use virtualization in a micro kernel 13:33.040 --> 13:39.280 by an operating system without actual implementing kind of an extra just using the basic principles 13:39.280 --> 13:47.040 that the operating system I mean the micro kernel already provides by using subhertz wonderful 13:48.000 --> 14:00.000 2018 I mean there were certainly more talks but this one stuck to my mind scout OS was introduced by the 14:00.000 --> 14:07.520 genode guys you know just to explain genode is an operating system framework so it's a set of components 14:07.520 --> 14:14.000 that you are supposed to take and build your custom operating system out of it while 14:14.400 --> 14:22.400 while scout OS is a distribution of that components so to speak which you can actually use as 14:23.200 --> 14:30.080 as a as your daily driver I don't want to take credit for this but I was really pushing 14:30.640 --> 14:37.200 pushing Norman to do this for many many years so I hope I at least contributing contributed to 14:37.200 --> 14:48.960 that idea so that you know he did it so that I shut up I don't know 2019 genode introduced 14:48.960 --> 14:55.200 a kernel agnostic virtual machine manager again on a technical level an amazing feed because 14:55.200 --> 15:01.280 if you compare the different interfaces that the different micro kernels provide for virtualization 15:01.360 --> 15:08.240 and making sure that you can have a single binary that supports all of them that's amazing 15:09.360 --> 15:18.160 unicarnals were getting stronger so there were several unicarnal talks unicraft OSV if I'm 15:18.480 --> 15:32.080 mistaken and others and again one reality check which shows that the nice of architecture is amazing 15:33.440 --> 15:42.160 unless you are dealing with buggy and useless hardware so since Alphoree needs to support 15:42.240 --> 15:51.600 real world computers with real world hardware world vulnerabilities we need to face some unpleasant 15:51.600 --> 16:00.320 mitigation efforts in the micro kernel itself to provide some end-to-end security and safety to our 16:00.320 --> 16:13.280 customers 2020 so that was almost world's largest super spreading event for COVID I mean I 16:13.280 --> 16:21.440 mean I wanted to present it as a joke but it was quite amazing that we narrowly escape this so 16:21.440 --> 16:29.760 that we narrowly escape this fate I wouldn't like for them to be in the textbooks for this 16:30.960 --> 16:39.120 but yeah also looking on the positive on the bright side of that for many people that was the last 16:41.360 --> 16:49.040 large event which they attended before we had to do the two or three year break 16:50.560 --> 16:58.560 that was M3 presented again a way how to take the micro kernel idea as we know it and push it 16:58.640 --> 17:05.120 even further push it into new grounds by incorporating hardware software code design 17:07.280 --> 17:13.520 there was this big resurrection of Nova so Udo was here presenting his new 17:14.400 --> 17:22.000 old new involvement in it and presenting the his the R&V8 port and more news to come 17:22.320 --> 17:31.280 and the next year this was my micro kernel room I'm not sure how yours looked like but you know 17:31.280 --> 17:39.120 this was the matrix chat this was the the presenter screen and this is you know the output 17:39.120 --> 17:46.880 the streaming that all the other participants saw and you know in the beginning you know 17:47.440 --> 17:53.760 for in the first half of the day it did not really work so well so yeah the infrastructure 17:53.760 --> 18:01.840 was under under stress but eventually it won't find and G note presented they're 18:01.840 --> 18:07.120 probably the device drivers it's one highlight from that year 18:07.920 --> 18:17.840 2022 the infrastructure got much better I think there were almost no glitches this time but we 18:17.840 --> 18:24.800 are still sitting in front of our machines at home or whenever wherever and G note presented 18:25.840 --> 18:32.480 G note lives presented G note on pine phone and there was yet another newcomer 18:32.640 --> 18:41.040 managarm fully asynchronous micro kernel based system which I'm still surprised how they managed 18:41.040 --> 18:47.760 to achieve such a large degree of Linux compatibility in such a short time but they really 18:47.760 --> 18:53.920 focused on that and it's an amazing feat and an interesting project to to follow 18:54.080 --> 19:05.360 and in 2023 we were back to Brussels finally and G note presented their new approach of 19:06.000 --> 19:12.800 reusing Linux drivers again this this is a topic that has been done maybe six maybe seven maybe 19:12.800 --> 19:19.600 ten times before in different ways but there was there there was still obviously a 19:19.600 --> 19:29.200 currently a possibility to do it in a better way for 11th time and this is what G note and 19:29.200 --> 19:36.960 other projects are using right now we have also known that Google Fuchsia is one of the 19:37.760 --> 19:43.680 most widely used micro kernels at least at that moment and there was yet another newcomer 19:44.320 --> 19:51.200 ours or helios this time let's just a new micro kernel but a new micro kernel implemented 19:51.200 --> 20:02.720 in a completely new systems programming language what more could you could you want 2024 this is 20:03.760 --> 20:10.640 this is something that I'm not really happy about somebody just decided that 200 and 20:11.600 --> 20:20.080 40 minutes of micro kernel talks should be enough for everyone so the length of the of the 20:20.080 --> 20:29.360 level has been shortened okay let's move on I don't want to say anything better about it 20:30.560 --> 20:36.720 now I fully embrace trusted and confidential computing that that year or at least that was presented here 20:36.880 --> 20:44.560 so again jumping on a new trend in the industry and G note presented there 20:45.440 --> 20:52.160 new build system or cross compilation build system which I think also is very very important and 20:52.160 --> 21:04.960 maybe useful for for you too and last year again a newcomer CMRX a new real-time embedded 21:04.960 --> 21:10.960 MMU less micro kernel built from grounds up and presenting new interesting ideas 21:12.400 --> 21:21.200 L4E went through some certification efforts which are some hard ways how to get into safety 21:21.200 --> 21:31.200 critical systems monogarm and redox was fighting some uphill battle with some old unique monsters like 21:32.160 --> 21:38.480 canceling for success calls and signals again something that most of the micro kernel systems 21:38.480 --> 21:47.040 eventually need to face because Linux compatibility is still important to many people and also something 21:47.040 --> 21:56.080 that nobody is looking forward to and G note also embraces trusted boot so that was the history 21:56.560 --> 22:03.440 this micro background that room today was amazing great selection of talks I will be reviewing 22:03.440 --> 22:08.800 most of them and so I think the future is bright thank you very much 22:17.040 --> 22:23.920 I guess there are no questions today so so have a safe trip home or enjoy Brussels if you still can 22:23.920 --> 22:26.160 and see you next year hopefully