 Hey guys, how you doing? Yeah, so my name is Josh. I come from Wellington, New Zealand to be my home there now for about six years About six years Got a bit of a Kiwi accent on me. So I hope for you can all understand that. Okay. I'll try and I'll try and speak slightly I'm told that Kiwi speak really fast So I'll try and speak but slower Yeah, so what else can I tell you? I work for Catalyst IT. There are sponsors here We've got branches in Wellington. It's head office Auckland Sydney has our office as well, and we have another office in England and We kind of do not just Drupal. We do Education so Moodle and Pertra, we do bespoke software. We run stuff.co.nz in New Zealand Which is the second biggest website in New Zealand next to Trademy, which is New Zealand's version of eBay And we we do a lot of stuff with telecommunications in the past and we do a lot of stuff with government being in Wellington as well So it's a kind of a big company. We're almost at about 200 people now worldwide And I work in the Drupal team, which is about probably about 40 of us worldwide So today I'm going to try and give you guys a case study on a project that we did Started actually back in 2011 and we just went live in August last year and it took a while and it was a Beat for us and so today I'm here to present to you guys and talk to you about what we did with Drupal and How we did it and some of the lessons and things that we've learned from it So I would want to start off actually before we go any further I wanted to find out by a show of hands what kind of people are here So could you raise your hand if you're a developer? Now the majority of you raise your hand if you're a business owner Put your hand up if you just get out of interest I didn't meet you Cool. I see that hand. Hey Site builder. Yeah, okay Cool first lesson. I don't think you'd be able to build this is a site But there's a lot of other reasons for that that's a cool. Thank you for coming Cool so a little bit of a background about scmp scmp they have been a newspaper company in Hong Kong since 1903 so it's a really old company They're English speaking even though they're based in Hong Kong They target the English majority in there and there's quite a strong presence of them there And they are incredibly paper focused kind of country Country company so when I first went to the scmp offices normally when I'm working at my desk I have You know my computer my laptop my cup of coffee and and that's really about it some headphones and a fairly You know clean desk these guys have stacks and stacks and stacks of paper Issues and issues and issues of magazines and newspapers all around them The offices are covered in paper. These guys are completely paper focused which was kind of a surprise to me I don't really think that the newspaper company would be like that But maybe that was just some sort of silly thinking of my part, but they had an incredibly strong paper culture Back in the times of the dot-com boom. They kind of got a little bit hurt as well So the whole dot-com boom happened and America and Hong Kong were charging out putting everything online and making it happen And they tried to go along with that culture as well and found that the model that they presented to the online web Didn't work for them and for a number of reasons They they kind of went a bit sour around online and so paper was still working for them And they continued on strong with that and so there was an awesome mentality that online isn't good And as we know that's kind of a silly thing Entility So we ended up winning at a tender they Went out and researched where they decided that they needed a new website. They decided that they needed to give this thing another crack they had our older website that was built on vignette and They literally could not hold traffic that they wanted so every day the site would be a slow hog during the morning traffic as it happens with media websites and When when real news came along and people started to the website it would just crash and That was really interesting because they actually have a business model in place whereby you can't view the actual article until you've paid to be a Subscriber so there are people who are actually coming to the website just to read the headlines and still crash in the website There wasn't that much traffic going through it So they realized that they needed something bigger and better to help them grow their online strategy as a company and They decided after looking around at different solutions on the web The Drupal was a good fit for them and they decided to go and find people in the Drupal community They could ask to build the website and so they looked and they found us They found a number of other companies and they asked them to tender for building their website in Hong Kong and They gave it to us Not to show the complete reasons. We do have as I mentioned of experience in building media portals back in New Zealand we together with Stuff that Kota NZ and Otago Daily Times probably do about 70% of the media online media in New Zealand with New Zealand Herald being the other player We also have a fairly large Drupal team for one of the bigger ones in New Zealand and we kind of considered the Enterprise guys in that space in New Zealand And we part ended up partnering with Rackspace in Hong Kong They wanted to have a solution that was hosted in Hong Kong and they could For performance reasons so we didn't have office there. We didn't have a data center there so we parted with them and We ended up creating relations together this photo here is a photo of us at the Media Awards in Malaysia, which was in December last year and SEMP took away Four golds and a bronze from the awards Okay So let's move into the challenges Firstly I should explain the next number of slides are going to have some pictures up on there behind the background These are all photos taken off the website Photographers and they have captions on the on the inside there which are completely unrelated to the Presentation they just there to explain the picture Okay, so challenges as I mentioned before SEMP had a incredibly powerful paper based culture and so They work in a print by design kind of mentality what that meant was they had software for the CCI news desk which would build their pages for the newspaper and then pushing that content out to other sources was an afterthought so paper was the priority and Then you would try and push that paper content out to other devices like onto a tablet or onto our website And that was a real challenge because a newspaper One single page got pushed out into XML and that one page could have one article It could have two articles could have five articles It could have half an article and a second article the second half of that article in another XML page it could also have like little teaser pieces that actually aren't real articles that you should ignore and There was also images and things that were associated with that which came sort of a zip file So it was incredibly hard to try and understand how papers work and how they don't work and How how we had to sort of pull that information and of course it went out of XML and They tried to describe the data as well as they could but journalists being journalists kind of decide to change things and they decide not to say You know this is supposed to be a byline text and instead that's what they put their name So that they're the author instead of putting it in the author field and things like that so it became incredibly difficult to understand and figure out things so there was a lot of Change in culture that happened at the same time to say journalists you have to use the system correctly We can't just sort of do it willy-nilly a Second challenge that we faced was scaling for viral news One thing about news media sites, especially when you're on dedicated hardware is that you need to build an Infrastructure that will last when you have viral events, you know, Michael Jackson dies or Earthquakes in Japan these big event everyone blocks the news sites And you have to make sure that you can deal with that amount of traffic But the infrastructure that you need for that is significantly more than when you're just operating in a sort of BAU format and so Historically companies have bought excessive amounts of hardware and try and make that work and then you end up just sucking away power as You know 80% of your infrastructure is not really being utilized so we had to figure out an infrastructure that was I Didn't have to use so much of the structure and it could scale and deal with the traffic Interesting thing when we launched scmp.com in August the Google analytics showed that the Mal of traffic instantly doubled the day that we launched and I don't think that was actually because It was so impressive that it went viral and we instantly doubled. I mean because it maintained that level permanently I think what ended up happening was that their system was so old and Couldn't handle their actual current amount of usage that for half the traffic They weren't actually getting surf pages when they were trying to go to the news website Because the system couldn't even handle it. So for the first time when we went live They could actually see and get a clear idea about how many people they are actually getting to their website and since then we've actually gone, I think 400 times over that amount of page views in a day and We haven't gotten alerted or being you know, there's been no real problems third challenge was paywalls Paywalls are kind of a big thing at the moment in news media they're trying to figure out ways that they can make newspapers work online and still make money out of them and so Paywalls everyone's looking at it space to see if that's the answer if that's gonna give them Traffic if it's not gonna give them traffic and figure out how how what's the right model for them So they have consultants who will come in and tell them this is the kind of paper you are This is the kind of paywall or the kind of commerce model that will work for your website And the way the ECMP were told to make their their paywall was to provide a kind of meter trial system So you'd see one article and it would be able to read it see the next article and a notification will come up Saying you have read two articles. You have six remaining this month And then you exit out of that and you read the next article and then you read two more articles You have no read four articles. You have two left remaining this month until you get to the end Sorry, you can't read any more. You need to subscribe ECMP. Here's how much it costs You know step to the next step and you can read it And the idea is to get people engaged and people enjoying that website and then they want to subscribe and For the most part it's work they have they have a lot of subscribers that come through and For people who are just sort of passes by they can use the site and not really notice that there's a paywall in front of it and they took that concept from New York Times and They didn't quite have the same budget as New York Times So New York Times paid I From from research they paid 25 million dollars for their paywall alone on the website They paid us $46,000 US To do the exact same thing in 40 days So we had a lot of challenges both from a project perspective And also figuring out how you can make this work without it Making it too easy for people to get around the paywall and the big lesson that I Think the client really had to learn understand is that if you want anonymous viewers to read your paper and apply a meter trial to anonymous users Well, you simply can't do it. Well, you can only just make it an obstacle obstacle for people to get around it So for those business owners out there who are thinking about doing a paywall If you want anonymous users to hit that paywall They we have to as developers implant something on their computer that makes gives them a marker, right? And that's usually a cookie or you could use flash or you could use local storage You have to have a way of identifying who that person is for the next time they come back But because you haven't asked them to have any investment in that cookie ie login Then they can clear that and you they just get as if they're a brand-new user so there's a problem there fundamentally with this concept which is Clear your local storage clear your cookies removed flash you get in for free. So that's still a fundamental flaw and problem You have to have users invest some way into that session make them want that session to be able to retain the amount of views That they have but they found it that it was just too much overhead to have people logging in that they would just Go away to leave the site if they had to do that and therefore We couldn't enforce it. They did however create a reward system sort of gamification if you will inside of Format so if you you could now I think now you can read three articles as anonymous user Then if you register you get an additional three for free and then they try and put the comments model on you So now you can sign in and once you sign in then we've got your email address and we can email you and say Well, why don't you do this and do that if you don't come back? We can send an email and offer some more free articles or something like that The other major challenge in this project was We've got a question. Yeah, so as I mentioned We implemented it and it works, but it still has that fundamental flaw So as long as being an anonymous user as if you know how we did it then you can get around it and That's that's the But I mean, that's unfortunately the case for New York Times and you know who's spent you know a magnitude of hundreds More money than we did and and it's still a fundamental flaw Didn't use the first three module. No, I do have another slide about using module Cool moving on So, yeah, the the next challenge was data migration As I said, they had a previous website. It didn't work out so well for them. They have been running since 1992 And so that was kind of a cool thing they'd actually been archiving data since the day they went online and That was something that they wanted to bring into the new system in the old system and their vignette system They could literally not have more than 10 days of news articles on the website and then it was too full So they had to every day purge the 11th day of content and they'd put that into an archived system and so they actually had this thing called the They were seven day index where that was like here's all the news on the side essentially You go through and read it and that was kind of how they they did things but it ended up turning into a feature What was originally a constraint turned into a feature and we ended up having to re-implement that for them Interesting, but what I also wanted in this new solution was to bring everything since 1992 into Drupal and that included over 800,000 nodes and 300,000 users and 85 gigabytes worth of data Which is incredibly hard to restore in your local environment when you want to you know Bring everything back up and try and work on a new feature because there's anyone tried restoring a you know 20 gigabyte my sequel database That's a nightmare because you know and if something goes wrong you've got to start again. I think it's four hours of your day so That was at trouble But that wasn't actually the hard part of the migration the hard part of the migration was things like subscription data Where you know between 1992 and 1998 they had this model and they use the database in this way And then after that date they had a new model, but they keep the same schema They just inserted the data in a different way so you end up with all these kind of conditions as you're trying to migrate data and and you're dealing with Basically corrupt data and figuring out how you can import it. And so the biggest problem that we had with with data migration was Exactly that it was that there was inconsistent data and that caused some problems questions on the back What database technique? No, right So the question was if I heard of a Particular module. I forgot my name already temporal databases And the answer is no I haven't and I didn't try that effort what we ended up doing was Moving all the data into a separate SQL database was actually a separate set of tables that weren't dribble eyes and then Because the other thing is has anyone tried using the dribble API things like node load and node save over 800,000 No, that takes a weekend. So you you know, and if it fails, you know, if you want testing It really isn't flexible. So we ended up writing SQL queries and statements So we had a bunch of we've got all the data into a SQL database Then we did manipulation to make the data consistent and then finally we had a set of SQL statements Which inserted it into the Google and that's kind of the way that we had to do it So we had to basically go through the back door Because it wasn't anything any other way that we could do it faster So there's just a lot of headaches because you know, we would go through Into a month into the data migration project and we go look at the starter. It doesn't look right And then you know the guy from the project team will pop his head up and go, oh, yeah That's because of this requirement that we didn't tell you about and and that's the kind of So I was going to give a demo of the site Perhaps we did a bit later on if we have some time Just want to move on to the lessons that we had so continuing on from data migration yet ten years of data is not consistent We literally had 99 problems logged in the data migration process and we had a guy in our team for Dell fixed all of them and It was a complete nightmare for him. I'm surprised he's still smiling So I would I would also recommend if you've got a lot of data like that Try and plan it as if you can get away with that a time of material thing because we found your nine-on-nine problems They weren't identified at the beginning of that phase of work They happened throughout the phase and it was at that stage you have to go through change requests and things And we actually ended up just going to time of materials because they realized that the data had issues And we couldn't possibly scope for things. We don't know about so it just worked better if we could work that way and once we had that Freedom then we weren't Constrained by time, you know, oh, we can't do this until it's been approved and all that kind of stuff So it moved much faster once we went to time of materials Face-to-face is better It's really awesome that we can work internationally and do things we have some advantages like We can work in the morning before they wake up and get down We can do deployments before they arrive so they can do a full day of testing and not wait for us to do a Deployment we've already done it by the time they come online We also leave early, which is not so good for them sometimes because they want us to continue working or doing something And we can also do deployments at their outage times not the non peak times and They're our prime time are working out. So it kind of worked quite well, but in terms of The project moving forward and moving at speed. We always move the fastest when we were together So time is when we went over to Hong Kong or they came over to New Zealand and that face-to-face environment kind of like you know If you're in an agile project and you're on the sprint room together It just moves faster because you can ask a question get the feedback get the approval that you need to go ahead and do it And you can think on on your feet Instead of having to analyze something put it in the email send it across log a ticket that kind of thing Becomes kind of tedious and you end up more in the prep project management space than you do and actually doing work in the project So face-to-face was always better. Oh, we also had things in the cultural aspect. They kind of changed things if you ever do work in Hong Kong The courting system over there is like like in New Zealand you sort of say here's our price that's that's what honestly it's going to take for us to do the work and They say okay, or no In Hong Kong they they say well, we're going to pay this price And they sort of counter you and and when I like so we put out our price like this is actually what takes and then they go Well, we want it for less. We're like, but that's what takes you know, and they're not they're not used to that so like we had to kind of learn this because If they pay what we originally said then they feel like they're getting ripped off and And if we if we lower our prices and we feel like we're getting ripped off So you figure out a way of like coming for a good agreement Yes We ended up going with discount models so We said look, this is our price They said no, we want pay less and we said This is our price, but we'll give you a discount And then the next time they understood that we discounted them for the last one And it's not going to happen for the next one And so we sort of mended the relationship that way rather than over quoting because it Can that's not really Yeah, we but in saying that come to the end of the project there was a lot of unforeseen things There was a lot of change requests log and we did come over budget to go over budget So we probably lost out on a bit, but in actual fact we still I mean we still made profit We're still successful in that Okay, so it listens no when to write a custom module was a big one We hit the ground running when dribble seven just came out and if everyone can remember back to when that happened There was still a lot of modules out there that weren't ready I mean they kind of had releases but they had bugs and that just kind of happens when you're using a brand new version of Drupal right because until people start using it people can't identify bugs So we can't fix that so we were willing to wear that we write when I we decided to contribute to that and help make things A bit more stable and we did that in the process But unfortunately along the way when we were viewed cast a contra module and decided to try and use it It often caught us cause more problems than it was with Views was one of them Unfortunately, I know this community really loves you so much so it's in triple eight But in this context it was actually bad and the main reason was because we were using views The way that you would use them in triple six, right? Which is you build your query and in the UI you place all the fields and you render all the fields independently And you don't use view modes or node templates because they kind of sucked in triple six But what they ended up happening is that while the query time could be kind of quick the actual rendering time would be really slow and we ended up with probably About 30 views on the front page Which is huge right no one has 30 views on the front page and when you're doing that many queries and that much rendering time It actually takes a really long time to get that out And so we had a lot of problems there and in hindsight would probably use something like view mode Which would be much faster an actual fact there were views was doing stuff like You want to render this field? Well, I'll load the entire entity before I render it, which is kind of important to me So is adding the into the cache module instantly improved performance because we weren't loading things over and over again We could just load them out straight out of cache But panels was pretty awesome So it was really lightweight We pretty much remove blocks and use panels and that meant that we were able to do Dynamic layouts for every single page and there was probably about 70 different templates that they handed over to us to build And so we did that all panels and we were able to consolidate a bunch of those templates into a single panel in it That was a great one. So would use those again and Would probably build some custom pains instead of using views in place of it Also another one module They wanted to do well one thing about paywalls and one thing about Having a subscriber model means that you have predominantly authenticated traffic in very little anonymous traffic and That's really important to understand when you're looking at it from a performance perspective Because in Drupal core in triple seven out of the box You can get varnish caching for anonymous traffic for free But you can't get that for authentic traffic as soon as there's a cookie involved with a session attached to it That becomes a problem That wasn't going to be good enough for us since the majority of our traffic was going to be authenticated We didn't want authenticated traffic to take the site down So instead we had to create creative ways of making Authenticated traffic Payable and that means doing things like abstracting the things that are specific to the user were like the hello username And putting that into something else like the JavaScript ESI or varnish or something like that And then that created additional problems Because there are other modules out there that would do things like creating sessions You know When you didn't want them to do it and it would slow down sites and stuff So we ended up having that single constraint in place meant that there was a lot of Contra modules out there that we couldn't no longer use because we were just poorly written And so we ended up rewriting a bunch of them to make the better But the maintainers didn't want to commit the patches because they were too big. So we sort of just roll alone Mega-menu mini panels There's an extension that allows you to embed a mini panel onto a Menu item and use that as the mega-menu for hover overs what it does is print renders all the mini panels in the page load and beds them at the bottom of the page and Then send you the page back now. I don't know if you've looked at media sites recently But they're all about really long pages. I want you to scroll forever So as it is you're already putting out a really large page now You're about to put panels markup at the bottom of it each menu item like five or six additional ones and render everything That's inside of it in there as well Which is kind of cool But the faster way to do that is actually to load them with JavaScript instead as a Ajax callback So that when you want to see it then it does it and you could actually do a late loading thing We did a lot of lazy loading. So You hover over it and if it's not already loaded because it loads in Ajax whether you want it or not It'll just give you a buffer and then it will show up and that was a much faster thing because not only you're not Spending the processing time and your page to to generate that But you're also not sending that down the pipe hole No, because that module exists and the way that we change the code is so significant We were that that patches wouldn't fly in that module Could you rip most of it out and rewrote it does sound like something that we could definitely contribute back in some space I've used it already in other projects locally. So it's a good point Yeah, you have a question the question for that was what other modules have we rewritten Another one and I wouldn't recommend using this module ever one of the requirements was Great way to do today one of the requirements was the user can only have one session at a time Think about this for a moment. So You log on to the website you can only have one session So that user ID that you ID can only be on that one device One of the old problems they had right is that pay for a subscription And then give the user and password out to all their friends and family and they can all log on to the site and use that subscription So what they want to do is lock it down to a single device But that is actually quite a challenge to do Because what happens when someone else tries to log on do you kick them off and let them go on or do you? deny that log in or you know What's what's the logic that happens there and then they also extended that and said actually you want to allow? Three because what when we when we launched people started complaining saying oh, but I've got a tablet I've got my mobile phone. I've got my desktop computer at home. I've got my work environment They're like, well, okay, we can use X number of those on on that website I suppose so they wanted to expand it But the the overall problem was that the session Logic required you to and the module we use was session limit required you to create a session Which sent a cookie which prevented caching and the way that we were doing caching and so it Basically meant that the user certainly logged in or even visited the website would just slow them down It became a real problem. So we ripped that module out and rewrote it I rewrote it. I don't remember what I did but it works Yeah, we should catch up later on We also did some cool stuff with chart beat. Does anyone use chart beat before? Chart beat is like If you've used Google analytics recently, you'll notice there's now a real-time widget board and you can see traffic hitting your site in Real-time chart beat does this feature only and it extends us using something called newsbeat and that's specifically targeted towards newspaper sites and so scmp they really wanted this this feature in there and they made us integrate it and Then we took a stance when it was when we were talking about analytics and things they want analytics and said We will not track analytical data because we've already got 85 gigs of data. We don't need anymore. So What we did instead is we were went back and query chart beat Query do analytics for that statistical data and showed that on the website So if you go to scmp.com and you scroll down the page and look at the most popular widget that's coming directly from chart beat in real-time and I originally had a bit of JavaScript in there that it would actually update on the page while you're on the page But I realized a it's not really that useful because it's not on it's underneath the fold most of the time So you can't really see it but it's changing and I also happen to crash the server the same day release that change Whoops Cool I'll move on to some stats that we have about our project. So it went from May 2011 to August 2012 We're still doing work with them and doing iterative releases, but it wasn't until August 2012 that scmp.com went live the new version of it We squashed over almost 1700 bugs So we had a bugzilla system set up and whenever there was a bug either from our QA team or from them that log it and would fix it Over 8,000 emails exchanged back and forth between the project managers 18,000 hours spent on the project 198 features delivered. So these aren't requirements not a hundred and ninety eight requirements 198 features which contain requirements inside of them that we worked through and deployed 70 pages of client documentation. What we did was We hardly ever used the book module So what we did is install the book module and built the book under the administration section of the site And that's in line documentation for the whole for the editorial team and for the project management team and the digital team themselves We did some a bit of training over in Hong Kong as well showed them how the CMS works on the back end and how they can run it 40 catalysts are involved in the project in some sort of capacity although there's probably about 12 core team members 10 trips to Hong Kong Three to eight hundred articles are imported daily So there's not just newspaper There's also magazines and newspaper and some newspapers are bigger than others. So that's why the number of articles vary We also bring in feeds from Allied Press and Bloomberg and these other news agencies as well Quite a few get commits in there And our infrastructure runs 21 virtual servers on six physical service Imagine some people might ask me why we used physical infrastructure instead of going to the cloud And the key one went there was that we signed a service level agreement with the CNP to guarantee performance of of the site and if you're running on cloud hardware You can't guarantee the resources underneath you So we wanted to be able to ensure it and if there's a problem that we could actually deal with it at the hardware level So we went with physical service and put our own virtual layer on top There's over a million lines of code that that we wrote But that is predominantly features and views and things like that. It sounds a lot more bigger than it really is There were a lot of modules that were enabled again That was a choice of how we decided to roll our features and a lot of those modules were feature modules And we did 255 releases to staging before we went live Pretty epic So that's the majority of My presentation like to open up the fourth question The question was there originally a print paper and when they go online. Do they keep the content from paper online as well? Okay like Okay, so How that works? How does the workflow process work in scmp? I think is All I can provide it as an answer which is they use CCI Newsgate that's their product for producing the paper and Journalists will have their own desktop. They will open up the software which is a desktop application and They will type in their story and send that to the editor the editor and his team are then responsible for choosing the I like course to create editing the stories and placing them in the paper They send out the typeset that exports to XML that XML gets sent over SFTP to us and we import those stories and they all sit unpublished in Drupal Then what happens is they have a digital editorial team that come on board They look at the articles and they start promoting some of them No cues to put them into the different sections that order them around the page and They published you know the published ones that they want and they will alter because what happens is something that's title and on an article and print is sometimes ambiguous Because I want you to read the article and find out more information if you put that on the web It's not what you're looking for so you won't find it So what they end up doing is they have a print head and a web head and so they'll change that around and Fix up grammar and things like that So they have a dedicated team for doing that and they're kind of responsible for they have the freedom I think to do that with the direction of the editor But they still responsible for doing that and they have a team of about 21 guys That put the paper together so that I think one thing that I do appreciate about scmp Was even because they are so paper focused. They do really care about content Which is someone who's a site border. I should probably know more about that But kind of I'm interested in the technicals and don't really care about the content They you know put a team that's bigger than us who built the site to be able to manage the content on it Which I thought was really cool They use CCI news desk the CCI news gate, which is a different product I feel like I'm selling them now they they actually and that's probably the better product They just is really expensive and licensing fees and they don't want to go down that path But that will actually export articles in XML format rather than use pipe papers into XML format Yeah, yeah, so journalists go there. They don't go on to Drupal so that it's not actually the authoring That's interesting I'd say if you go to Drupal 8 that sounds good because you can actually have the offering environment on your phone But in Drupal 7 unless you can get an admin in the face that does responsive Okay It's not a multi-lingual site the sc&p even though they're in Hong Kong actually are only English paper and anything that's in Chinese Mandarin It's actually usually an image because it's something that they're advertising No So what they yeah back when we started actually I don't need responsible that's the around so it wasn't in the design And so when you go to it on your iPad you'll find that you'll get the desktop website And if you go to it on your phone, you'll get the mobile website in the comm and So that's the strategy at the moment in the computer commas incredibly Compact down to headlines and actually bear it because it's much smaller and works a bit faster And I think on Android it goes to him that is to be calm. They are currently working on Educated app to download and work It's yeah changes a lot with these guys. They don't care so much about accessibility. They care more about Proprietary ownership so because because it's so important to hide content from people it makes it more less Less accessible. So for example, we have the paywall that will pop up and high content and there's interesting things like If you use I think Apple has a reader type device where you can choose to extract content out of a website It doesn't support mobile and present that in a way That's a bit easier to read and understand and that was for a while a loophole to abstract the content and not go through the paywall So we had to figure out ways of blocking that out and then there's you know, all these kind of crazy things You end up having to Change the content and make it hard to access unless you access it in a way that you can control Which means you can't send accessibility out the door We looked at open publish. It was in triple six at the time and we didn't want to build on triple six because It's not future proofing though. By the time we release they get a year of support So that was a immediate no and the size of the project really meant that it needed to be Spoke Should I be integrating with Union pay as a commerce gateway No, so they wanted to integrate with a payment gateway called Asia pay and in Hong Kong and Or pay dollar was there their product that we use and so we we use the commerce module and Drupal and You know built out all of the products and things and integrated that into it and then we built a custom pay dollar integration module to That was the only payment gateway that they wanted Did we get any extra support and from the government or for some other entity to help the relationship in the project? Not for scmp.com But there was a trade and enterprise group. I can't remember if that was government or not that we're interested in and helping us out develop our Business in the Asian market and so we are currently pursuing Options I'm not I'm not working at that level. So I can't say but I think so. Yeah We're continuing the relationship. So it must mean it's something worthwhile Yeah, there's a lot of Data fees that go in and out of Drupal site And yeah, they've had this system in place for a very long time I remember obviously there are papers since 1903 and they've moved from having everything on file to into a computer system And they have all they already have their infrastructure set up and dealing with paper-based subscription And so they wanted us to integrate into that and the way that they've been doing it in the past and the way they wanted to continue doing that was through synchronizing data feed which was Terrible Yeah, yeah, they use Oracle finance is one of them. So if we if we sold a Print and digital package together We'll give them the access on the website and then there will be a feed that would at nighttime run and Send everything from the day over and then reconciled that and then everything over asked that had other ways of getting to the Digital site and would reconcile that and yeah, so you run into race condition problems And what happens if someone does something over here and over here and all that kind of stuff We just have to kind of deal with it. It's not a prettiest solution, but Well, thank you all for coming and I hope that's been useful to you We'll just call it there and come and see me if you have any questions. Oh before I do finish We have a boss session around digital media And that's happening after this at one o'clock So if you're interested in that come along and Okay, what time is that? Okay, how do you feel about that David? Yeah, unfortunately, I'm leaving this afternoon Yeah, so we could come to your talk about paywalls. Okay, so we'll go do that as well and Until and it's a 10 year old room and until 1 o'clock Before her talk, well, I'm free and happy to talk to anyone else about anything else around media Performance scaling paywalls commerce all that stuff. Thank you