 Good evening. Now we are going to have a talk about Deleon web service development with Frank Lichtenstein. Frank? Lichtenheld. Lichtenheld. Lichtenheld. Excuse me. Okay. So this is ended as a buff. So obviously I will not talk the whole hour. In spring my plan was to actually have something to show you at this time. We live in Tereent and I have not. So today I can only present you some ideas what we maybe should work on and what might be worthwhile goals and encourage you to come to me and work with me or present me your ideas or work alone and make it better than me and actually produce something. So I have neither a code and only a little plan and with some of the stuff I will talk about I have not really much experience. So I would like to come to me and share your experience with me so that we can together create something really cool for the Deleon project. So one thing I want to talk about is the fact that we have many, many websites and many, many services and there's a whole lot of ways to interact with them and not all of them are obvious. Some of them are very mysterious, not good documented and they almost all do not share any information about yourself. For some people that might be a good thing so if you do not want to share your information with these websites, that's perfectly fine and we should not discourage that but I myself would sometimes prefer I could just give some preferences to one of these sites and have the others some way to learn about that. As I said in my talk description in the schedule it's not my plan to adopt Launchpad for Deleon or something like that. I mean Deleon doesn't work that way and we have a historically grown farm of services and websites and that's certainly something we do not have to abandon but some things are currently guessed often and stuff like that that could be set by the user instead and so to have reliable information for example the mapping of which developer which keys belong to which developer which different email addresses belong to the same developer stuff like that. And we have also some things that rely on only half automated tasks or something like that like for the website we have some of these things that could benefit from a central place where users and interested parties could insert that information. So obviously we have in Deleon I will explain some of these things because they are written very short so obviously we have our LDAPS accounts as Deleon developers and obviously that they contain some information that is private and that should not be exposed over any web page and has not be to accessible at any web page and it has the problem that it's for DD's strictly only so there is no way and probably there might be no good way the moment to include users and DMs and stuff like that there and you have the mail and the web interface to change these values the DDPO is one of the examples where I really would like to tell the site more about me at some point and which supports something like that over the mail interface or via the cookie setting but for the cookies setting it obviously works on your own one machine and I have to set it again if I change the browser or something and for the mail setting that's not at the moment I think authenticated anyway or stuff like that so then for the Wiki I have a separate account that's again and then for RT most for the request checking system for the admins and stuff most of us don't have an account but if you have one separate account again and then for alias yet another account and even though some of the group stuff I think was still synced with the LDAP database but only in one direction and then you have your subscription to the list but okay and then you have the BTS where we in the BTS booth offered some people requested some features which would require something like authentication to the BTS something like an account like login but at the moment it hasn't so at the moment it doesn't need to right so I think that should be clear so one of the ideas I wanted to present here and I will spend most of my time on doing that is that it would be nice if we could have something like a common identity that gives us the possibility to tell all the different websites who we are and even maybe but that has not been the first step to store preferences in some kind of database wherever that might be that depends on the technical solution for the problem so that other sites can ask about that if we order them to so for example all the email addresses I use all of them are mine and I should and I'm more interested in all of them it would be nice if for example the BTS would know that because the BTS is probably one of the cases where most people have due to some circumstances use different email addresses but some people use even different email addresses for their packages key ideas packages I'm interested in for whatever reason stuff like that or for example am I a user do I need much help am I a developer and I want to be the display to be concise and very much information and little explanation stuff like that I would find that nice so one idea one could have to implement that would be an open ID or some similar protocol for HTTP there are some smaller implementations like the stuff the cpa the retest tracker for cpa uses this bit card account but they now support open ID too so one idea could be to encourage to first encourage Debian web service to support open ID lock ins and secondly to help that effort by setting up an open ID provider for Debian users that don't have one otherwise so that we could easily support some of the more advanced stuff which might not be supported by all providers and which also could mitigate some of the disadvantages of open ID for example that in itself it provides no trust because you can normally use any provider and so that would be a provider we could trust more for example because you could authenticate by a GPT key or something like that so open ID the basic way it works is that you authenticate to your open ID provider which can be yourself which can be a company that offers that and which then tells the server the web service the application question that it has authenticated you and you can then map this open ID lock in to some account and in the newer versions of the protocol there are even some nice ways to exchange preferences so that you could at some for example at some central place in your own provider create a list of packages you're interested in and the services could ask for them and even provide a push update facility the problem with open ID is that to my knowledge at least works only for HTTP based services because it was developed for them so no luck for your SSH lock in for other stuff right and of course you have to take care of actually encrypting the lock in and stuff like that yourself so another way would be to use a more low level approach a more a general approach like for example Kavoros or something but that requires a lot of setup which would would have to be done mostly by DSA so it's not a lightweight solution and I would think we would have something like that if it would be in the interest of DSA I mean it's it's not an obvious solution the other problem is it's not where we will integrate it in most end user programs and stuff like that I think for for most browsers you need to plug in or it's not supported but as I said at the beginning I don't have really much experience so if I talk completely about it please tell me so and one of the nice things of OpenID is that everyone can provide their own provider so nobody has to rely on someone else for their authentication okay so that would be one thing but if you have something else you want to talk about in terms of what could we do so that all these websites that we have all these services that we have don't are so nebulous at the moment they don't look alike you don't use them alike mostly you basically have to know them so you can find them and it's not a very integrated approach ideas would be welcome so most of the service you mentioned actually have content which is fully public and not really user specific so just to understand your idea you're thinking about using this kind of authentication to save preferences basically like filters and this kind of stuff right? yeah I think as you said many of these services currently do not have any login but I mean it would be nice for the services that have login to support something that would work with the solution so for example someone could look into how much work it would be for alias for example to support OpenID plugins and I think some of the services we at that at the moment don't support any kind of login could profit from it but and I think if we if we agree on a solution beforehand we can make less work for everyone involved so for example if we want to add it to the BTS the BTS and package Steven Ork if we work together as the maintainers of these services we can make it less work for any of us but of course you have to be interested in it at all had you thought about using an LDAP database for storing that because that's what LDAP is about I mean not our LDAP database for the development developers but an LDAP database is exactly and it supports more than all the tools for the ones that you need PAM, you can use PAM and for web services most of the languages have plugins for that so even you can publish part of the tree as public I mean if your preference could be queried by the applications without any login and that could work yeah right so what you propose is to set up a separate database as the source how do I say to the service which user I am no you still need an account on that LDAP database I mean for DDPO for instance we can tell DDPO to query the login I mean for this example you can say oh query for your account and get the other email address and make it all together and then we can just keep adjusting the the fields or the object classes inside the LDAP database to provide information for other things like probably BTS if you need something different like it needs to have an login page because for the DDPO I think it's easier because you already know the mail is just a matter of having something that connect then and then queries the LDAP database would be easier for other things that needs to restore the preferences then probably it needs some login and then it too needs to query the database probably you would add some load to the SCA if you are going to have it as an official service but it could be started as some web services.debin.net or something that we can like try a few things and see if it works yeah I mean the the advantage of something like OpenID would be that it would be even more decentralized so you could provide an OpenID provider but nobody has to use it if they don't want to but yeah it's of course nice idea too so yeah basically I think it amounts somewhat to someone starting to do something I mean the problem with all this stuff is I mean it's not the first talk I hold in that direction and at the moment there seem to be very few people and Debian who work on that and those who work on their websites on their services are often very busy with just maintaining them I mean the BTS has made some great progress in the last years and packages has quite some changes and the BTS has seen quite some changes but there seems to be always not enough motivation and time left to do something which could benefit all of us so do we have an appropriate forum or a place where we can discuss this kind of stuff besides the single main of the people involved that's a good idea yeah I mean we could for example create an adage project for that and create a mailing list there and discuss it there because it's not really on topic on each of the single mailing lists like Debian QA and stuff like that so yeah if enough people are interested in we could definitely do that and use that or maybe just Debian website mailing list yeah we could try that too but the point is the extra traffic because I don't really care about WWW Debian or I mean yeah actually I think the problem of the Debian WWW WWW mailing list is that it has quite a few types of traffic that are very different and I often find it difficult to follow it and ignore all this stuff I certainly would prefer an own forum for that but only if more than two people are interested in so okay I mean I can set up an adage project and I send it over the appropriate lists and everyone who's interested can subscribe and we can talk about it more after Debian I would send it to Debian QA, Debian WWW and Debian Conf discuss probably so yeah yeah some other things that could be on topic there would be for example more yeah machine friendly interfaces to our websites like the BTS has this SOAP thingy which seems to be now used quite often from different groups and teams and stuff and something like that could certainly be added to more of our websites that would be something we could discuss that maybe too so other actually the SOAP interface for the BTS is being developed actually I think that for most of the service can be added really simply and I think that having a SOAP interface to all the various services can really help people in mixing data together because it will be just the same technology yeah and maybe if we add SOAP interface at some point for the UDD then we have all the data yeah yeah we had a quick talk about this topic we all think I guess that's the websites of Dubian are very heterogeneous very different looking as long as it's on dubian.net I don't worry too much and it's no problem but we probably all feel that dubian.org should be more similar looking I'm really interested in getting the wiki dubian.org having something similar to dubian.org even though currently we all feel like it's not very nice looking so someone I can't remember just said we could just ping some graphic websites and having a design I should ping this guy we can work on having nicer design that's one part yeah the problem in the past as I feel it was often that there were some design proposals for the website for example and stuff like that and it often then faded away I don't know exactly what the problem was but it seems that it was missing some dubian developer really taking into I think we have a problem in that area to accept contributions and to encourage contributions so we say that we want a new design but if someone turns up and has a new design we seem to fail to go through with that I've seen many proposals in the past coming to the mailing list or dubian on the wiki there was a layout proposal and I think if we call for people to propose a new design we really have to help or maybe just have the design and implement the design ourselves because I think in some way it's just the place where it just got stoked we had a design and yeah the guy couldn't implement it or just couldn't find the person to help implementing it I don't know if you share the impression that it's really at the css the time where the css had to be done that it's stoked so there was a layout but it was hacked and not really usable in the form and had to be implemented really with some of the proposals there was really the problem that they were just hacked together images and there was no really code behind it and someone would actually had to make the crude work to create the templates for that would make it look like these images yeah okay yeah we would obviously need someone that does css better than me for that but yeah right well could we just have a design contest and then that way we could just kind of have one theme that some people pick and vote on it and then we could just apply that to all of the websites that need applying I mean that seems like the most efficient way to have a single design that everyone likes and people get into those I mean they sort of like the recognition and stuff like that and it's kind of fun for the artists and designers to get together and work on stuff and feel like their contributions matter and at least with the wiki there are multiple themes so it's not like there would even necessarily be a loser you could integrate the top five themes or something like that into the wiki that the users could use yeah I mean I think it generally only takes someone to organize it to lead that effort I think we really have to assist the designer to implement it because maybe it doesn't but quite a few websites and CSS would have to be re-implemented for each websites more or less yeah right but you can have one shared CSS that will cover most of the websites and the HTML should I mean the HTML for the frame should probably mostly constant and then if you have a one website that really looks like it and is implemented in HTML and CSS and is not a generated image then it's mostly trivial should be mostly trivial to apply that to the other sites and because that's mostly just one work then but for me at least the difficult step is from the Photoshop GIMP image to the HTML plus CSS I mean obviously we would need to make it the context should be about the HTML and CSS and not about some image yeah but you really you do a CSS file once and you just have to put in the style information or the style classes into the general workflow of the system you are using so you really do the design only once yeah but I think he meant that you have the existing format of each website and you would obviously there would be some adaptions needed but but one point that I want to aim at Erin it's not the first time that the contest was proposed and it was never really opposed and it just a contest will definitely work and no one really objected to it so far but nobody was several times proposed so far and no one really organized it in the end behind you will there be some sort of a reward for the competition or what was it okay I mean if we talk to the DPL we could probably do something but I don't know it would be probably up to the organizer of the contest to decide that and to talk to the DPL what's possible and what's reasonable I have no idea what's reasonable for something like that if the recognition is enough or if you have to give some small price or I don't know that should be possible at least yeah the question is who should organize it should it be DB and WWW or the I think we have a publicity team I think to say that at least should do something never works I think someone needs to step up step up and say I will organize this and I will talk with people and I will make it happen and unless someone does this yeah and I think it's unimportant if that person is from from the web team from the publicity team or from some other team or not in some team at all it would probably help if it's DB and developer but yeah and it will be judged by who, by some committee or how will we work out the best yeah I think it was before my time how was voted on the logo that was a GR right might be a bit overkill maybe a simple a more simple web based vote would be enough we can use Mako's Mako has an election tool that we could use in theory I think but let's not get too hung up on that that's you know if we get so busy talking about how to do it it will never get done yeah I think that's all valid questions that should be decided but do not actually need to be decided here and now whoever decides to actually run it gets to decide how we decide it so I mean whoever wants to put the work makes that decision I mean obviously some people should actually say beforehand that they will accept the decision and the voting would have to be probably so reasonable that I will happily do that but in the end the technical details are up to the organizer of the contest okay yeah I will happily implement the results but I'm not the designer and don't won't have anything to do with that before it is decided right so I think the topic is some other topic related to any websites or stuff like that that you want to discuss we have time but we don't need to use it if we just sit around another one not very related to web services I apologize about this it's about it's very close to the idea of Frank of making it easier to switch from one page I mean if you're on the package for the package dash it should be easy to move to the bug relative to dash and so on I was thinking it should be not very difficult that the header of all websites that for instance are connected to the package to the binary package have links to the two or three other websites related to the website to the binary package so for instance package jbn.org, qa.org bugs jbn.org so the idea is generally to move the existing links that are between all these pages to one place that's identical for each page so that you can just take two it looks similar and maybe add some if some army thing I'm not sure I haven't analyzed this very closely but same for source package same for main dinner of the package and same for submitters doesn't matter I don't think so yeah just submit a patch against each of these sites and we will happily apply it then I think we're done for now and I will create the LH project for the web services stuff and then we can talk some more after that then thank you all for coming and let's get on with that