 I don't think it's a good start, I think it's the same part of it, but I think it's a different thing or it's a hard part, but I often comment between this management and QA, that this management is actually to create, get into a package in a good state, you know, in a digital access table state, whereas this management is releasing access table packages. It sounds a bit different, but in the end, in my experience, it's just two different ways to look at the same thing. So that's one of the most important things I learned, and actually if you look at the release team members, all of them are very active in QA also, and more or less you can style it as of now, more or less I've written the choice of the release team to be able to create as everything, no matter in uploads or whatever, because you probably need to know more than just your own package. That's a common thing, and that's why I was trying to release QA, so that we should talk a bit about what we're doing with the release, and also what we planned to implement for the next version for Edge. So, yeah, so if you look back, what's happened well and not so well, we see a lot of things where we couldn't improve, I think. Of course we see a lot of things that happened well, but that's more for the talk of more audience, whether they were all good or not, now we say we can improve more. So, yeah, one thing is that it was not working as well as it is now, it's outside now to memorise itself, so it has to be an end to all of these processes, but oh, it's good to know why it's an extra force thing, so now we're done. That's good. So, now let's see if we can get to my slides. Sometimes you see volume problems much better. Problems will solve themselves if you ignore that problem enough. Yeah, but sometimes it's getting worse and worse. I would be better if I just paid for this. That was a long, long time ago. Yeah, I think it was long after everybody leaves, but that's not the way that things start anymore. You ignore a big problem and you get much useless. Hey! Can we see some slides? Yeah, we also have a local line to go to. Pardon? Yeah. Don't have me, I know that we can have one of those extra lovers. Send them to me. Send them to me. I don't mind. We'll just back to the system up. So, we see the next one. I think we have here my last one. So, do one of you know offhand how to make full screen mode? In GB. I know for XPBF. Yes. Yes. It doesn't affect full screen mode. Now let's see if it's there. And it's true A. Yeah. There's something like that. Yes. It's something like that. Oh, I think it's the FD. Yeah. Oh, yeah. So, now let's see how it goes. So, okay, now it's my first slide. Yeah. So, that's what I already said. So, the next slide is what we have. Now I have something like that. Okay, look up. No, you should look on screen. Not that bad. It's not that bad. It's not that bad. It's not that bad. Sorry. Sorry. So, what about the rally mode? It's money storage. So, if it's very simple. They got, shimmy tools, bugsp advisor tools. You every lightange. Of course based on this was my Twitter account. I think that's not a filter for ZTE Plugin. But in the end it is not very fine. ZTE and Z ez updates. updates that we will announce was very good. Even if we didn't know it was things already, this time that if on the end we will do that, and then we can do a date end plus these days that allowed us to start learning even for a time that we didn't know was infrastructure really ready. So eventually it was fine. One thing that is definitely important to be aware of is a new policy and basically for more than half a year a new policy was if it's broken upload a fixed version now, which is not the official any new policy before was if it really needs to be uploaded later, at least to the later in the late four plus fourteen days. And for a very long time we worked with if it's too broken you can just upload a trunk to unchecked so that you can store it on the same day. And that's something you need to consider if you want to which any new policy you want to continue. Of course some, as I see back things that still upload in the late seven because the upload is a little bit more careful so I'm checking if it's a maintenance or more time. Most as I see back things it's so simple you just should upload there. And that is actually what we are currently doing for the C++ API transition. I personally think we should go to a more aggressive and a new policy. Of course it doesn't mean just upload anything you just like but if it's broken, it's a maintenance and also brokenness you should be able to fix it and you don't have to wait for the maintainers own personal good mood to do it. Also what we are well, at least my opinion of course is the management of the release in large set of course starts at the time where I looked first into the release team as if we are at the first release announcement where we are at the time where I was at the development of the song I don't want to comment about December two thousandth of the announcement. Yeah, a lot of the songs were done in large and had a small set is read our, basically read our release announcement. Yeah, I don't know a lot of things that are not meant for the releasing. Take a look what's the transition sound. The transition sound is one of the stuff that I would come back later on because it sounds like I have been very much to this game. Yeah, a lot of more things could have come better but I want to highlight from the season art is this what we need to reconcile should we also release some development packages earlier? I can put all the slides in the back of the phone. Because for example somebody managed to upload a CDBS fix to unstable just five days before the release which was like that some packages had a CDBS untill then very very long. I would say that CDBS is now at the back of the release. But it's not too good. Yeah, so that's it. A lot of small stuff but it was a cute situation, wasn't happy at all. So what we want to do in Edge is we want to do a timeless release in December 2006. We have said we want to not so much, we don't want to make as many release blockers as before. Only again a small stuff set and the rest of the important thing is more of an A-like approach. We want to do it but we don't manage to finish it. It's still okay. We can live with that. But what we definitely want to do is we want less dependencies between libraries and A-lit bumps. That makes life hard for the release team because it plays a testing position. It makes all the life hard for the security team. So those of you who were here during the security discussion, the members said, oh it's really bad. If I do upload of a new version ofこちら, it will be like this. If you upload 17, it will be like Scallion and so on. So that's something which we should do as QA. Of course we want to do less pain with some arch but there's now my opinion which is probably on the way because for some M6 AK performs now better than the previous M6 AK. So in this example, one can motivate people to do their own QA on the arch texture. That's also something that should be left for QA to report us. Yes. And lots of more small things that we want to do. Of course there's some ongoing QA which has happened already before the release of Sage. It should happen now, it is happening now. That's when they had a full-fledged task because I was just normal and I was maintenance of QA off the loop. Patches need to be build able in testing and in unstable. That sounds very easy but it's quite often broken sometimes later. Then the next is ongoing QA works and patched on the build in P-Builder and if it doesn't build anymore a C-Paxon file, that helps very much. Something else I should put down by the maintenance actually is that it checks the upload before they put it on front of the P-Master. Yeah, patch question is just normal QA work but it helps very much for the release process. There won't be a release if there won't be a patch question practice. So that's very, very, very well. And more guys on this for the release process, I think that's something I think a QA should just say we have a package and we apply some guy. But we should really say how can we help to maintenance to make this thing better? Because it's case better so it's more fun for all of us. It's just better than my opinion. There's something new QA is that we want to do. We want to work as lively manifest stuff. So Steve is doing some work with lively manifest stuff. It has two things in mind. One is to know better which stuff the library is actually used. And the other is to make sure that the library is a simple version at the moment. That's something that would help in a lot of different releases. Second one is we want to get rid of at least some of the specifically compiled libraries. We know for example that Zlip is directly compiled not only in such things like open office or de-packaged, but a lot of other packages and modes are even unknown. So that's something simply to say you want to know where it's in and then fix it. And the next question is does a package is possible to install a package or move it again? Zlip is doing some work as a pure part. Zlip is working on the last item that's in my opinion quite important is actual pricing policy. Check it. Because it's really broken if we have more than two different versions of pricing in the same release. It's the same as of course for other things like more than two different versions of Zlip, new TLS or our five different versions of the version database or similar things. Yes, I have a painful, not only for the release team for the QA. It's something that really seems to be where we, you know, this meeting says we want to have some, but we are going to last let go by dying out of this policy company. But at first the first goal is QA to work and if we as it seems out already the disk is in the box but we just don't notice it because we don't check for it. Or we will say we start first with QA and in the end we say there's only a few left and we believe it will become the disk. So what we want to do at the edge edge, that means die edge for the disk, we want to check the installers of these databases before we accomplish it. I think we all learn that with less than very hard. So that's possible to play if you can do work as carefully as you want, but you should really do a check with the final version of what the STDase and the final version will look like over this time. So this is the final release of the final and official Archer's date. And then from the official Archer's date may somebody just place some release files by hand and forgot something. That was better than happy. So that was the release notes, should start earlier. There was a lot of stuff on the release notes. One of the things that we started is that we have now absolute package for the release notes. So if anybody makes a change to the effect on extra release notes, this file now also won't fit. That sounds a little simple, but it will definitely help very much. I already put a bug. I've seen it. I guess I should make a bug as long as you were fast as I need. Yeah, so you see, we can learn what is important to the phrase to get an experience, to learn an experience, to put an experience for other people. That's one of the important things. There were now some slides about how to create the release interface. I think if you ask a question I'm happy to answer the sentence and discuss them. I think you have some rather first question before to ask a question. Yeah, a few slides before. A few slides before. The first slide with the negative points. Now let's see how he can go back. Hi, Mr. Pepp. Is this one or the other one? The other one. The other one? Yep. The point was don't drop that packages not for the next stable release to Unstable. So that means as all some packages in currently Unstable which have a series of wave back which just says don't put this package into stable. For example, TMEA. Yes. Or there was another page I can't remember. And doesn't mean that such packages go back to experimental. Well, I think the other side is a bit more difficult. If you have a package that sets a leave of the package I don't mind what. If there are testing, there are packages I don't really care. But I care for example if there are not so many packages called for example KMEA. Then we say one to release with the version 3.2 and two days before the team's position happens. One day before the main channel uploads the first part of KMEA for the main channel. The first part is very important to emphasize that KMEA team has a group better than much. At that time I said KMEA sucks so much but I know it's good. But now I want to say now it's a very good KMEA team. And now we're at the team and I really have this KMEA. I think we're almost done. At that time is such a thing. And now I want to answer because I understand anyways it was an incredible maze. So the set is not acceptable. Of course it is a police package so I don't mind. I think the answer to the question is a little bit similar. Our seed box on the packages are on packages that are not testing and will not be released at all. And that point is about packages that will be released and KMEA in the meantime starts an upload before the first one. I see. So it's more about don't upload new releases than don't upload packages. Yes of course. It's a maze. The answer is I understand what you're saying but I hope so when the next answer is in the second chapter it will be but I got my answer. That's a slightly different story. I have a question about the test process on the CD images built for the next release. I am a little bit concerned that checks being done on the CD images which is the very last thing before the release was done as well as they were done this time. Do you plan to to have automatic checks like this in the English language? The problem system was something else. The problem is what happens this time? I think the full time of how the release happened. So the release happened in the morning about the last there was an extra the release was just a lot of last as you can see backfitted from the in some hand you can see there was extra the release and after the release the middle team started to the last middle the place by hand some parts in the middle the release files but not by the release files started to build CD images from it CD images installed as a studio of course but they had some security line because as a studio line was at that time checked out if someone testing it actually the best idea was to commit something like four weeks before the release or two months before the release the security they wouldn't need any issue but what would have been better at that time they both know the least version of that they were able to check and on the next day the simple things were changed so the release started to build the middle team the CD image team the other one was the real one and for the next time the plan basically my idea is basically there are different opinions of course but the least opinion was what is down to if you make the last file then the strict even though it's not a problem then we change the sibling makes the release and a better few selected people can have some FTP master so these people have the official strings and a bunch of people who are doing a lot of quick errors say this dude installed a computer and then we take about not only something like actually we had something but also changes to these we hope to make something like a few days of changes to these so we had too much further than the last days if you want to add something on your point another point is when we talked about WNCD earlier and WNCD has a lot of places where you need to configure things to use the proper archive to set proper things and they were just a couple very very minor mistakes made at that time because of time pressure and WNCD will be rewritten before release so that should make it a lot easier another thing is that we will probably switch to using the proper release names instead of the code names instead of the face names and that will also get rid of these problems because the install will then just put in charge in the testing security line instead of stable or testing I would however still suggest that there would be to prepare and keep current a kind of checklist to be done a list of checks to be done against city images there can be really real test cycles for the city run so that definitely it would be very nice to only hit a single button to build the cities after the city team is working on that and also we will make much more check this, we will when everything there are a lot of mistakes that we can make the same mistake twice list up to stable mistake the check except we won't do the mistake I was going to ask out of the city if you know how there is city deviancd rewriting is going because we should keep an eye on that so that it doesn't become the edge blocker just send a mail to deviancd list and Steve McIntyre will hopefully give you an answer but he really presented his plans quite extensively there were a couple of sessions held about it and he even intends to go so far as to rewrite the city writing tools because there is stupid licensing issue with them currently and we have this crazy upstreaming maintainer for them but actually there will be Sunday before the before the release we will say all shows are done by now come we done we will make check we will take care of that but yeah when of course Lanzler just said mark is one downside between the day we have like last week the public release they won't be in the store anymore they will be in the store I saw the movie of Louisville I think that what was just described is very important because the release so far has been mere eccentric if you like and we should go to a more overall approach where you take both mirrors and CDs and the release notes and everything else that goes with release into account there was the pressure that it is released same thing for setting up the security infrastructure properly within the time frame you also have the time to do that same thing for maybe maybe even do last minutes changes to the bootstrap and make sure that the new testing is supported instead of having complete inability to install testing yeah and so on there are a lot of small things that we can do better but for that we really need to say we have one week on every time to make the final small adjustments but I know that one of the step to P-Masters very much is that when these CDs are being pushed they are simply some of the caches that we definitely want to have one FTP master has said well it is the image we are going to work in time and just publish a sibling and we are done which is something I personally would rather not just do so what about just setting the sibling and then wait for a week for fixes to come and you can still push stuff into the problem that you usually will be able to say and then declare the release as soon as the sibling is published so what we intend to do is we put the sibling in place but only allow selected developers who we are sure that they won't leak out to excellent it when you can't leak it out as long as the end user doesn't see it then no one actually gets it so maybe allow everyone to see the released version and you just say maybe we are going to do changes and then you can publish the finished play of that series and then you can and Monday when we are sitting on this Monday afternoon I said this time we are going to Monday evening and this time we are going to push everything but actually on Monday morning I want to hear one host name one idea is of CDimages which is normally not published and Monday morning the man is brought out which means you have already half of our full uploads of these and people start to release these in a full in place because it was last time something like 15 minutes you would be challenged to as in high-technology at the moment that the sibling is visited to public there is no way out we can have one mirror that will get in the news start first you have to make your point and you don't you don't promise that is the final thing you say this is our release camera but people are going to listen you should stop you should stop you get this last story you don't even know that but that is part of the release story sorry what do you mean by release camera it is available on it my life, my death experience what will happen if we put it on a very hidden place it is somehow available to the public so first when it is released on it there will be no some mirror operators who will just use very little mirrors so fdp can't stop and pull from that mirror it was more current then we can't if you make a part of the normal release process so it will just be like the current testing or edge or whatever it is called and people don't rush for that it is the release what happens if there is an important problem with that people have it sold and you cannot provide one more to the final release go to hit decent shit ok one thing the importance what we are doing in the testing is that we will be free so people can test but what we really need we don't need an almost release thing but it is a full release thing for the last test cycle and we need to be able to repeat this last test cycle is full release stuff as often until it is now the final in nothing changes so it is a full release thing it is a full release thing looks so full release so we can play it out that is not for the set it simply pops up and it is broken and I can remember on the other leg of the release where all people dip up a sponge up here and then be in good and then be in the middle so it is nearly half of the middle which was before the release announced that I can and basically but my experience with management of the release is there is no way that 30C is it I am breaking you can stop it I just wanted to add the same thing you are pointing on at the moment you are pointing to the wall happened to free BSE about 2 or 3 times do you mean it is a full release or what? they push the release up to the mirrors without having the announcement out and 15 minutes later there is a slash drop the people are starting down the ring and the free BSE people had big problems saying we have not any release announced as long as this hasn't happened this is not the release they just didn't manage it the only thing they managed is to 2nd or 3rd time they managed to get the slash drop that far they didn't bring that size yes but that was the only point actually when our plan was to smash it which happened out of my relevancy was we made top level mirror pushes to family from the mirror in the US in Europe and Asia and we know that the mirror push is an important part of the mirror push takes about 1 hour till it is released so we said ok we will set it out of the release and also wanted to have our 8th push start at the end and we set it out of the release to have time for the release because we are so sick so we have to move it to the end normally and so on one question when the CDs are tested and the mirror pulls ready I believe that the images have to be built in each principle mirror so the images are for the CD images we have two things for the CD images one is the chip to find and the other is small and we have CD images CD image network is the most part of the CD image network is an important part of the CD image post so what is the set what the e-book will do is to know long enough in advance they could push images but themselves so as they are in the middle they just need to be moved and then we need to say at this time we are going to publish and they are going to log into each other and make this move to the host and then it is public and a mirror will have to do that the CDs and the the CD images and the CD images are actually produced at almost the same time there is no time lagging that the only thing that lags behind is the turns and I thought that each printable mirror instead of syncing the images they pull the GTO files and build the images you know I remember I just finished a mirror and I would have to ask there is one thing that is important there is a chip to find actually in the e-book because it is the first e-book and it is not so good to collect the site so at that site the images are built the chip to find are generated and the image that we work they are the chip to find the CD images are very important I mean but the CD images are very helpful and if we say in advance we will make this process they will be able to contribute to us ok, something else I would like to go on or spend some time on is about this whole library of transition stuff one thing that is really very helpful for us currently is a transition of different libraries for example the Lib Check one they made a library from the SO name ODOT 80 to ODOT 100 yeah, it is all the package was removed at that time but about 30 to 40 those packages are depending on that package and that means that all of these packages needs to be fixed at the same time we go on to testing and this is not only true for that but even worse if there is some, for some reasons they need to implement an action version into a stable later on because I am bug fixed all of these packages need to be built in stable which is really painful so for testing there is some easy idea to solve it so that it is just to keep all of one of the loading packages in testing as long as as long as necessary as long as necessary until all packages disappear but for stable it can't do that for stable how does it work for stable if for example Lib Check is an easy one but say for example Lib Cairo is going to the same way for Lib Cairo we are creating here a norm all depends on and as there are some security bug fixes there will be some new Lib Cairo to fix it that bounds this domain but the other is a bug and that's something which would really hit us with more than that but it would be even worse if you would have such a security issue in SQL that would bump it's not only about bump this in no LibGCC LibGCC I didn't want to hear that because LibGCC I can take a bug from a potato and then I can LibGCC not necessarily not the other way around I can take a bug from Sahaj which is a C yes but the other way around is something like API compatibility thanks to library version but the possibility of security fix and you need to break forward compatibility it's something like highway highway do you mean like a silo? it does happen in the kernel it does happen in the kernel it has it I mean in a number what's the number of reverb I don't simply need to come to a clever solution but I think that's something that we need to keep in mind some other life sentence so any more questions would it be possible to make it easier for developers to see what packages are involved in major transitions that are going on at the moment I know the information is available but it's not very accessible actually currently it's so bad that I wish I would know better what's potential involved it's really bad because the C++ API transition is so large it's really bad if it's simply that high for example I know how many people really know that the next KDE version is tied into the C++ API transition because of course it's technically okay but it makes it hard for outsiders to understand what is going on there well developers or users well for users we send out mail before use we say it's unstable but one of these users suggested that there could be an announcement for all users that could get this kind of stuff so this has to be such that it's a code that we don't announce no I'm going to send a mail to DDA with the user details about the transition if you have this upgrade you will have to remove some packages so you don't have to do that I think there should maybe be a page where you'll have these transitions are currently by the release team that are being tracked we want to move them to testing so we don't want new uploads and you'll have a field where you can enter your package name and you can see is this package affected by the transition for an old package it's quite easy to know yeah that makes sense it's very easy you have to do a huge reverse dependency check uh... that's all you have to know it's not a package I think I know how to spend this I think that the training has been really high but as a user point of view it's much better I just discovered that one of my packages was actually containing some C++ files which I actually forgot because if you have that code available you could make a check in a product processing and say sorry, this package is not amount at this point because of transition going on this is a cool idea I think I know some people who like to send a C++ check but it's not a cool idea to make such a page okay I know so thanks for all