Plurchase Blog

Shop with Friends

Delayed launch, but more features!

By Dean

We were asked to delay our launch, but there's no sense in waiting around!  In the past few days we've added features that beta users were asking for, but were too big for us to fit them in the initial launch.  We've added private browsing when users enter into a page protected by SSL.  This notifies that their own connection is encrypted by SSL, as well as preventing other shoppers in the same chat room to see what you're looking at.  The picture shows what a secure page looks like.

Another feature users have been asking for is support for multiple tabs.  Many of us browse the web normally by opening up a link in a new tab so that we can look at it later.  It is a way to remind ourselves that we want to take a look, but "not right now".  Previously you could still open up browser windows in new tabs, but you wouldn't get the app bar that contains the chat room and other features.  We took a look at how GMail does their GTalk chat system and we're trying to make it equally similar so that shoppers will feel right at home.

Stay tuned for our real launch early next week!

Filed under  //

Comments [0]

Questions answered

By Dean

Some of our technically minded beta testers had some questions about security and privacy of the system.  I thought it might be good to mention what actually happens on the server side to let our users feel more comfortable using Plurchase.

  • We don't log messages in our chat system (the box in the lower right corner).  What you say to your friends is none of our business.  The chat messages are sent to our server, then rebroadcasted.  
  • When you enter information in web forms on a merchant site, this data is immediately proxied to the merchant, we don't log any of this data.  If our server is ever compromised (even if they steal the physical server!), the thieves won't have access to any data about purchases you may have made through our shopping system.
  • If the merchant uses SSL on a particular web page, we will also provide that same webpage using our SSL certificate so that your data is encrypted with the same level of security that the merchant originally intended.
  • If you enter data into web forms, users who are shopping with you will not see what you typed in.  We're not a web conferencing solution -- your shopping state is completely independent from anyone else who may be shopping with you.
And finally, we'd like to borrow Google's motto "don't be evil".  We're here to create an entertaining shopping experience, not sell customer data to evil people.

Filed under  //

Comments [0]

Our Tech

By Dean

I figured when we hit front page Hacker News, people will want to know about some of our tech.  Some people may not realize how much technology goes into something simple like Plurchase until they do a right-click view-source.  We've got a wonderful extensible custom proxy server, integrated with a real-time comet system, with lots and lots of javascript glue code.  When we talked with PG, he kept pushing us to work harder and get it launched before the end of the funding round, but when you've got all this technology under the hood, it's hard to make things work right.  If Plurchase was done in 1999 when the web didn't have iframes, flash, and ajax, we would have been done a few months ago.  We probably still have a few more months of hard work to make the shopping system work for all merchants, but we're quite happy with what we've got so far.  It's a MVP that users will actually find valuable and a product we can feel happy about.  The technologies we picked will enable us to make lots of quick iterations on the core product to create features users will love.  

  • Ruby - technically JRuby/Rails.  Ruby has a great community of hackers, even if there people like _why and zed that make it interesting (btw, they're great people if you meet them in person).  Stuff like Facebooker, OAuth, and Rspec make it extremely easy to implement new features in a very short amount of time.
  • Javascript - mostly JQuery.  We considered using Sproutcore and Cappuccino, but we really liked the jQuery framework for doing UI work.  Most of our MVC stuff is home grown though.  Personally, I'm hoping Sproutcore finalizes their 1.0 base soon because the last time I looked at their alpha versions, they were basically unusable for production work (about 3 months ago).  They use jQuery as a core library so our migration there will likely be easier.  Makes sense to use a framework here soon, but we only have around 12,000 lines of javascript so far.  We've also implemented some awesome stuff that made code fast:  monomorphic inline caches in javascript (non-jquery) sped up some parts of our code by almost 20x.
  • Java - people like to say bad things about Java.  There aren't many features in the Java language, but at least it's decently fast and has lots of great libraries.  We have 94 jar files in our lib directory, but some of these we aren't making full use of yet (like shindig & nekohtml).  It's interesting to use both Hibernate and ActiveRecord at the same time.  If you look closely enough, you might notice where GWT and flagstone are being used.  
  • Comet - specifically Jetty/Cometd.  We're slightly concerned about its ability to scale, but hopefully when it becomes a serious problem, that should mean we can hire someone to help us improve it.  Supposedly Jetty's Cometd can handle around 10,000 concurrent long polling sessions, but we'll have to see just how likely that is.  I figure we won't even hit 100 concurrent users after launch.
  • Haproxy - nothing really fancy here, but it's a life saver when you're trying to integrate comet, rails, and servlets under one hood.  We're using stunnel for ssl, and shorewall for the firewall

Filed under  //

Comments [1]

Only three stores?

By Dean

We're only a day or so away from launch and some of the beta users are wondering why we've only got 3 stores right now (in beta mode, they could see hundreds of stores).  Adding a store is easy for us -- it's just adding 2 rows to our database and then setting a visibility flag on it.  We're launching with 3 stores to keep our product quality high and so that we can find out roughly what types of stores others are interested in.  We've heard from a number of beta users that they'd love to use Plurchase for buying hotels & plane tickets.  Apparently it's actually pretty hard to coordinate travel when all you've got is email an instant messaging.  We've also heard from a vocal minority that they want to use it for Ebay auctions -- lots of bargain hunters like to scour craigslist and ebay for goods and compare notes on product lists.  

The female crowd is clamoring for a set of boutique online clothing stores -- most of which I've never heard of.  We've added a little UserVoice widget on the search panel so that users can vote on what stores they want to see and we'll slowly add them as we verify each store for correctness.  Majority of the sites work fairly well, the ones that are harder to get right usually have lots and lots of javascript & flash -- sites like orbitz, kayak, and bing travel might require a bit more thinking on our part.  

A few users also complained that we didn't support their web browser -- IE6.  We're planning to support it, however it's not a great use of time right now.  We want to focus on improving quality and adding new features for shoppers.  If you're stuck on the IE6 web browser, we do provide the option of installing the IE6 Google Chrome Frame plugin.  I've used this on my old Windows XP machine and it works great.  Simply attempt to use Plurchase and you'll be redirected to a page where it will prompt you to install the frame plugin.  Once installed, you'll have to close all Internet Explorer windows that might still be open.  The next time you visit Plurchase, you'll have the Google Chrome experience embedded into your IE6 web browser.  Also, if you're using IE7, this is currently the only option of using Plurchase.  Although IE7 is marginally better than IE6, it's still not a very good web browser compared to the rest of the competition out there.

Filed under  //

Comments [0]

2 more days until Launch

By Dean

I can't believe how close we are to removing the password protected homepage.  The hardest part of launching is convincing the other founders that the product is "good enough" for a public release.  There were a few show stopper bugs that caused some of us to pull all-nighters to fix, but they have been mostly resolved.  There are still performance issues and a few javascript bugs that may reduce the shopping experience, but if we have to destroy every single last bug, it may take us until December to launch.  We want a quality product, but we also know that it takes a long time to build a community of early adopters so hopefully those of you who use Plurchase in the early days can bear with us while we pull all-nighters to fix your bugs and user experience warts.

Over the past 2 months, I've pulled almost 20 all-nighters at the Hacker Dojo.  Some of our bugs are caused by regressions in the code, and others from not testing a specific browser under a specific platform well enough -- we really need to get our buildbot/selenium system running so that we don't have to manually test this stuff over and over again.  Our tests consists of successfully shopping on various merchant sites as well as running the Mozilla Firefox test suite, jQuery's test suite, Mootools, YUI, and our own JsSpec tests.  One day we'll also have to incorporate Adobe Flash tests as well, but most of the big name merchants don't make heavy use of flash yet.  Our mix of languages requires us to run everything from Java TestNG tests, Ruby RSpec, to various javascript framework/library tests.  Without a real automated testing system, it would probably require a developer at least 30 minutes setting up each test environment and running tests.  At one point we considered adding Clojure as one of our core languages, that probably would have made things even more brittle.

Filed under  //

Comments [0]

4 more days till launch!

 

By Dean

It's interesting how nobody cares that you exist during the stealth mode pre-launch period.  We've got zero comments on all of our blog articles and we've only got 3 followers on Twitter.  If you do a google search for Plurchase, this blog is one of the first hits -- nobody has ever written an article about us.  I can shutdown our production server and nobody will care.  We've been holding off implementing lots of really cool features because we want to establish our core product first.  In 4 days, all this is about to change, and I'm really looking forward to it.

 

Filed under  //

Comments [0]

Hard at work

 

By Dean

We're only a few days away from our selected launch date and I can't believe how many bugs have been reported and subsequently destroyed.  I've been spending most of hacking time at a great community center called the Hacker Dojo.  It's really a wonderful place to work.  All the members there are friendly and hacking inspired -- so much better for motivation than spending solo time at a coffee shop or in my own apartment.  The Hacker Dojo is technically a hackerspace; a place where people can meet and work on their own projects.  For $100 a month, I basically get office space that is open almost 24/7, as well as the ability to host any event of my liking.  A friend of mine hosted an epic birthday party here a few weeks ago.  They're trying to keep the events hacker-centric as a way to attract other hackers to become members of the Dojo.  The Dojo is located one block from the main Y Combinator office and roughly half mile from downtown Mountain View, with bicycle lanes that connect directly to the Dojo building -- perfect for anyone who wants VTA or Caltrain access.

Filed under  //

Comments [0]

Simple problem, hard problem

By Dean

Allowing people to shop together online has always been something people cared about but nobody has been able to solve well.  How could such a simple problem be that hard?  When we tell people about the technology that is used to run Plurchase, we blow their mind every single time.  Most of them are just astounded how much technology goes into something as simple as allowing people to shop together online.  

Some competitors are trying the embedded widget model.  They develop a shopping widget that would go on the merchant site and require a partnership.  This is something that we had considered doing at one point, but after talking with small merchants, we realized that forming partnerships would take a very long time, and most of these companies were worried about anything going on their main webpage -- a lot of them were not technically savvy.  It would even be harder doing it with a medium/large sized merchant since they wouldn't see the value in spending development effort in it.  

Other competitors have tried doing the download model: "install this plugin and ask your friends to install it too".  The previous incarnation of OneRiot attempted that, along with dozens of Firefox plugins, "social" web browsers and the like.  In a previous life, I was a Firefox plugin developer so I definitely could see how easy it would be if we just did it as a Firefox plugin.  If we did the download model, we'd have to develop plugins for every web browser & platform out there -- not an easy task since there isn't a sole browser with the dominant market share.  

This leads to our current strategy, the proxy model.  It is significantly harder and does cause some sites to break in certain places, but we're rapidly fixing every new bug we find.  Social shopping is a simple problem but has a hard problem to solve.  Any time you find yourself writing a javascript parser in javascript for the purposes of something simple like social shopping, you realize that some of the simplest things that we use actually has lots of technology inside.  For example, the iPhone -- deceptively simple to use, but has lots of technology underneath.

 

Filed under  //

Comments [0]

Startup products we use

By Dean

The Rackspace Cloud - They power our development, test, and production machines.  So far they've been very dependable, I only wish that we could have a 4GB instance that only had 10GB of disk space - the default of 160GB is too much and probably the reason they don't support backups for instances of that size.  

UserVoice - I first saw their product at a SHDH almost a year ago and instantly fell in love with their concept but I didn't really have a consumer level product to make use of it.  We're hoping that our users can make good use of that tool and tell us which merchants they like to shop at.  I did have to make some hand modifications to their default javascript file though -- they've got document.write() embedded in there which makes it impossible for us to load the script asynchronously.

Mixpanel - That product is kickass -- I'd recommend it to anyone who wants to do real time tracking of how users use the product.  We've got our own A/B testing and funnel metrics framework integrated with their stuff.  

Google Apps - Most people take their stuff for granted, but without their apps, we'd probably be using some POP mail server with spamassassin.  It makes me really happy when I don't need to do sysadmin work to get basic stuff like mail working.  

Posterous - Owning a blog couldn't be easier.  Not having to deal with Wordpress security updates every few months is also a huge plus.  

Filed under  //

Comments [0]

Which one should I get?

By Dean

We struggled to get a decent user interface that was both easy to use and useful.  I think we've got something pretty decent, but we're still looking for ideas to make it better and easier for people to compare products with.  If you would like to see something improved in our interface, please drop us a comment below or send us an email at feedback@plurchase.com.  

When you get an email from your friends with a small list of links, do your friends usually put a price tag at each link, or do they force you to visit all the links and discover for yourself?  So far there hasn't been a great way other than email and instant messaging for doing the group comparison/commenting/shopping.  There's the Facebook wall to help with that, but I think the Facebook wall moves too quickly for a persistent place to comment and rate a small selection of products you're thinking of buying.  

Currently, you can mark pages for comparison & discussion with your friends by clicking the yellow star next to each product image on the purple Plurchase bar.  If you've already marked the page before, it will simply jump to the marked location in the product wall.  

Below you can see our product wall that will let you compare and discuss buying decisions.  You can enlist your friends via an email invitation, or recruit them by posting the product link to your Facebook wall.  When they visit the link, you'll see them in the chat room interface.  Please let us know if there's something additional you'd like to see in the product summary.  Also keep in mind that this "product wall" can spam across multiple online retailers.  You can find something on craigslist and put it on this list when you press the yellow star on the plurchase bar.  Everyone else in the chat room can see your product wall and add comments to them.

   
Click here to download:
Which_one_should_I_get.zip (93 KB)

Filed under  //

Comments [0]