Multithreaded

Here are my thoughts regarding Google Chrome and the impending OS war with Microsoft.

Firstly, when Google first announced the Chrome browser, I immediately saw it as their OS play against Microsoft. Chrome has a markedly different architecture from other browsers. Each tab and window runs in a separate process, which isolates the effect of a problem in any one tab bringing down the browser along with all the other tabs. Other browsers still run all tabs in one process, with multithreading for performance. This means if, say, the javascript in one tab hangs then the whole browser will hang with the potential result that the user kills the browser and all it’s tabs.

What is running in those tabs? More and more it is applications, not least google applications including mail, docs and, soon, wave. Using Safari, Firefox or IE at the moment if one of those apps crashes then bye bye browser and all your other apps. This, of course, was how a lot of operating systems behaved until relatively recently (at least those not based on Unix or similar ideas). The dreaded blue screen of death, made famous by Windows 3 and Windows 95, meant losing all running applications along with all their unsaved data. Although our operating systems are now much more robust, as more and more apps move into the web then these sorts of crashes are re-emerging as a function of the browser. Chrome, the browser, overcomes this by letting you kill one tab (application) whilst not affecting any of the others. Great progress, indeed.

So let’s consider the question of what exactly an operating system is. Since my early days of reading Tannenbaum, through studying Unix and various other operating systems, I was exposed to the geek view of the OS which is a kernel with a collection of device drivers, memory handling, process switching and control and all the other bells and whistles which go on under the hood to make it possible to run your applications. But there’s the key thing. From the user’s point of view the only thing the operating system is there to do is to enable them to run their applications. So as the applications move out onto the web, the cloud, the grid or whatever you want to call it, all the user needs is a way to get to them - hence the browser becomes the operating system in a very real way, and Chrome, with its multi-process model, is well placed to be the leader in that direction. By taking the browser and putting it on the most minimal of linux distros they can provide a system which boots quickly and runs cleanly. If course, not all applications are suitable for running over the web like this at the moment. Although there are some basic graphic and video editing sites already, for example, I doubt that any Photoshop or Final Cut power user is going to find something that provides what they need in a browser. It remains to be seen if and how that might change.

The key foundation of this strategy is, of course, the idea that internet access will become all pervasive. Clearly we’re not there yet, but the trend is certainly in that direction. Google also mitigate this need with Google Gears, a way of building web apps which can work offline and then sync with the server when net access if available. It’s not ideal, but it helps.

So how are Microsoft responding to this? A large part of Microsoft’s revenue is derived from the relentless upgrade cycle of their Office suite. This has served them well over the years but may now have become a millstone around their neck. Any attempts to move away from Office tends to result in them being punished in the stock market as nervous investors struggle to see how they would replace that revenue. Recently they have announced that they will be producing Office Live, basically putting their office apps online and offering them for free, as Google do. This is a very risky proposition for Microsoft. The key thing to remember here is that Google does not make money from software. Google are an advertising company and that is how they derive in excess of 90% of their money. They can afford to build apps online and give them away and then wait until web access and browser technology catches up with their strategy because these applications are further conduits for advertising. Microsoft, on the other hand, by giving away their online apps will simply eat into their primary revenue stream. Personally, I don’t think it looks good for them.

So having looked at what Chrome OS is (or can be) the big question I see is why? Why are Google building a browser and an OS? After all, their apps can already run in current browsers (a feat assisted by using the Google Web Toolkit to work around the vagaries of the javascript implementations in different browsers), so why build their own? I can see two reasons for this, one trivial and one more meaningful. The trivial reason is because they can. Google is populated by a lot of very smart people who are given a lot of freedom to do things they simply think are cool, and re-inventing the browser must seem like a very cool thing to do. They are also driven, it appears, by a hatred of Microsoft. This, of course, is a dangerous motivation and has led to the downfall of more than one company (e.g. Sun, Netscape). It is important that Google not be blinded by this hatred and wind up the same way.

The other reason I can see is that having their own browser gives Google control of the whole user experience of their applications. This is drawing on Apple’s playbook but potentially mixing it with a strategy that Microsoft drew on heavily a decade ago - embrace and extend. Microsoft responded to the emergence of open standards and protocols by adopting them and then making proprietary extensions which would lock people into their own implementations. In Google’s case, this might take the form of applications which work well in any browser, but better in their own. There is one particular way I can see this happening. I mentioned GWT above. This is a toolkit, developed by Google and freely distributed, which lets a developer create user interfaces in Java and then translates those interfaces to javascript which gets sent to the browser. One of the key things about javascript, though, is that it is still relatively slow to execute compared to native applications, even taking into account the new generation of javascript engines. What if Google were to extend Chrome with a more efficient GWT runtime engine. GWT apps could then be translated to a form to run on the Chrome engine much faster and, possibly, with much more sophisticated UI features and interaction with the user’s local operating system, and better communication with the server. The apps would be good on any browser, but better on Chrome.

So there you have it, Google Chrome as a stable, multi-process platform for running web apps, especially Google web apps, and pushing Microsoft towards a position where they have to give away their main revenue stream in an effort to compete. No wonder Microsoft are trying so hard to break into the online advertising space to replace that revenue.

Development continues apace on our new application, Bamboo. What is Bamboo? It’s a web app which aims to provide support for lightweight business processes for agile organisations. Pretty much all organisations, no matter how small, evenutally reach a point where they need basic processes to manage things. With Bamboo, we hope that they can implement these processes with a minimum of effort and time so that they can get on and do the things they like to do.

We hope to launch in 6 to 8 weeks with a basic site offerring one core feature, Checklists. Checklists let you capture the steps of a process, say creating a new build of your software or handling all the tasks for a new employee starting, and then filling out a new copy of the Checklist when you need to do that process. Once you’re done you can file it away for future reference. How do Checklists differ from todo lists? Firstly, a Checklist is for when you have to do the same procedure many times. Secondly, the result of completing the checklist is always there for you to go back and look at in case you need to audit the procedure.

As well as developing the app, we’re also using CrowdSpring to help us to prepare the design. If you want to see the current contenders for the logo then have a look at our project.

So this week we took the plunge and started developing our first Ruby on Rails application. As long-time Java developers this was a big step, and I have to admit to being very happy. I’ve been avoiding RoR for a long time, mainly because of being put off by the level of hype surrounding it, but having come up with an idea for an application a couple of weeks ago I decided to give it a go.

We’ve made very good progress in one week. We’ve got the core of the functionality implemented and I’ve managed to get a mongrel cluster up and running behind an Apache load balancer. It certainly lives up to its reputation for rapid development.

Still to do is all of the design, payment merchent integration and all the usual behind-the-scenes stuff but I reckon we should be ready to soft-launch in six weeks at this rate.

The application itself will be a subscription-based service, very much along the lines of the 37signals apps but with functionality which complements, and will eventually integrate with, their offerings. I’ve been very surprised by the whole ecosystem of basecamp-like applications out there. We’re using Basecamp and Highrise ourselves, along with Beanstalk for subversion repository management. The biggest surprise, however, has been the number of screencasts available on how to build stuff. I especially like the Railscasts which are usually 2 to 5 minutes long and just cover one really useful tip each.

What does our app do? Well, the short description is “Painless Processes” and the slightly longer tagline is “Lightweight business processes for the agile organisation”. Almost every orgranisation, no matter how small, eventually needs some kind of processes just to keep track of things. The aim of our application will be to support those but keep them as light and unobtrusive as possible.

I just bought a Drobo to make use of a bunch of old drives I have lying around. I’m a big fan of redundant storage systems but my recent experiences of using an Apple RAID card in my MacPro have been less than satisfactory. The performance was not as good as I’d expected, it takes ages for raid arrays to initialize and I was getting a lot of problems with the system not coming out of sleep so I decided to ditch that.

I ordered the Drobo from Expansys on Thursday and it arrived on Monday. Today I pulled it out of the box and set it up. Firstly, the unboxing is quite a nice experience in itself, comparable to opening an Apple product. The all-black interior of the box, complete with black foam and a good quality black fabric bag on the Drobo itself, all lend an air of quality to the experience. There is a bit, three-step, getting started guide on the cardboard cover and all the cables and accessories are laid out in a separate tray.

So what of the device itself? The unit is smaller than I was expecting, just large enough to take four drives. It has nicely rounded corners and a very solid feel without beeing excessively heave. The front panel is held on by magnets and is removed with a gentle pull to reveal the drive slots. These have spring-loaded flaps on them and inserting drives is simply a matter of sliding them in until they click under the restraining clips. The Drobo takes any combination of SATA1 and SATA2 3.5″ drives. I put in a 750G, two 500Gs and a 250G.

The Resource CD has tools for both OS X and Windows. I was installing on a Mac which was just a matter of running the installer program for the Drobo Dashboard. The Dashboard provides several useful views of the device configuration from the overview:

to the more detailed status:

Following installation, the software offered to upgrade itself to the latest version, which it did with no fuss. Then it asked me to plug in the Drobo USB cable and it set it all up automatically, only pausing to ask if I wanted FAT or Mac OS Extended format. Within about two minutes I was up and running with 1.1TB of free space, all fully redundant spread across the 2TB of drives (1.8TB actually usable). This is far more impressive than the half-hour or more it usually takes to fully initialize a RAID array.

All in all, I’m very impressed with the device and the setup. The next test is how well it performs. Clearly the USB 2.0 connection will be a limiting factor here, but I’ll keep an eye on it to see how it goes.

Update:
Mind you, given this announcement, perhaps my timing for buying one was not so great.

I’ve just copied my system drive over to a new hard disk and re-booted from the new image and suddenly Eclipse will no longer start up. With some experimenting, I’ve found that using -Xmx768M was causing the problem. If I reduce this to -Xmx512M then it runs just fine.

The thing is, that the only change to the system was booting from a new drive. There’s plenty of memory free but it seems that the JVM is no longer able to allocate more than 512M to the heap. Go figure.

Just been listening to the March 14th newscast from the JavaPosse where they spend a lot of, very interesting, time talking about the new SDK on the iPhone and the apparent restrictions in the terms and conditions which prevent people from distributing apps which run on a VM. This would mean that developers could not distribute Java applications which required the VM to be present to run. They discuss a lot of interesting points, and speculate a lot on the motivations, sinister or otherwise, of Apple in deciding to limit developers. Obviously, not being Apple, they’re not able to draw a lot of definite conclusions.

I’d like to offer a different perspective on this issue. As developers, we see the iPhone as a development platform with huge potential. Since the moment it was released, I’ve been wanting to have other apps running on it. It makes a great connected device and it’s easy to compare it with other, open, platforms such as the Nokia N800. However, I believe Apple did not create the iPhone, and are not looking at it, as a device for developers. This is a consumer device, first and foremost. Apple’s priority at all times is that someone who has bough an iPhone, as a consumer, should get the best possible experience all of the time and if that means that developers are left frustrated then so be it.

I think that the directive against VM-based apps it based on this principle. Apple do not want people to come browsing the software on the iPhone store, seeing something they like and then being told they need to download one or more other components in order to get it to run. Or worse, downloading (and perhaps paying for) an app only to find it won’t run because they didn’t know that they needed to download (and possibly pay for) another component before it would run. Supporting this in the iPhone store would mean dealing with dependencies between components and then you get into all the problems of version handling - what happens if I have the Java 5 runtime installed but another app needs Java 6?

Looking at my experience of using a Nokia N800, it was great that it is a very open device and that there are lots of third-party apps for it, but my experience of installing those apps was often frustrating as I needed to locate libraries that those apps depended on and download them separately. Even as a programmer I found it very annoying and tedious. I can well see that Apple to not want that to be the experience of the average user who expects things to just work.

In fact, I would not be surprised if Apple did not take “it just works” as their yardstick for deciding how and when to open up aspects of the iPhone. As the JavaPosse discussed, the delay in releasing the SDK was probably not down to not wanting people to develop for the iPhone, but rather taking the time to ensure that the experience the developers have is a great one which probably meant considerable cleanup of the libraries and tools between the launch of the phone and the release of the SDK. For iPhone users they are going to be even more stringent in applying that principle.

Of course, another aspect of that is that they would want all apps on the iPhone to be as consistent as possible in their look and behaviour. The first part of that is the visual component library which means getting people to use the native SDK as much as possible. The second part is the UI guidelines, but Apple developers are usually pretty good at sticking to those. I have used Java apps on Symbian 60 phones and generally they look and behave very differently from the native S60 apps which can cause a lot of confusion and just be downright annoying sometimes. Again, a situation Apple would love to avoid on the iPhone.

In conclusion, I think that if you take a step back from your developer viewpoint and look at the iPhone SDK and software distribution system from the point of view of a consumer then Apple’s behaviour seems a lot more understandable and a lot less sinister.

Given my rants about the customer service, or lack of it, I received from Orange back in 2006 when I made the mistake of taking up their broadband service, I thought I might redress the balance a bit by singing the praises of two companies I have dealt with, one recently and one longer-term, whose service is exceptional and diametrically opposite to Orange’s efforts.

I’ve recently moved house and of course I needed to sort out the broadband. I looked at some of the more well-known offerings from Virgin, Sky and the like (funnily enough I didn’t consider Orange) but then I remembered a recommendation from a friend many years ago - Andrews and Arnold. Look at their website and you’ll see it has none of the glossy, flashy marketing stuff that appears on most service providers’ sites. However, behind that site are a number of experienced, intelligent, dedicated techies. Okay, it’s a little confusing to find the information about pricing and bandwidth, but it’s very encouraging that right on the front page they mention things like Reverse DNS and IP6. They even have a link to an IRC channel right there on the front page!

Now admittedly, your average householder looking for a bit of broadband for the weekend is probably going to take one look and head off to BT or the like - this is not a service for them. For the likes of me, a long-term techie who is as comfortable configuring up his own firewall and setting up SSH ports, etc, this tells me that these people speak my language and know what I am talking about. They are also refreshingly honest, telling you up front that the bandwidth you’ll get is dependent very much on the quality of your line and distance from the exchange and that they’re not going to make promises about that until the end of the first ten days by which time they’ll have monitored the line enough to tell what is the optimum speed to drive it at.

But the real kicker was when, on my way to work two days after I installed the service, I got a text message informing me that my line had just dropped. I immediately called the support line and got through to James who had all the details of the line history, noted that the line had been dropping off-and-on over the past couple of days and suggested that I might want to take the front panel off the phone socket and plug the router directly into the BT master socket to eliminate the internal wiring from the circuit and so see if that was causing any problem. It would also ensure that during the line training period I’d get the best quality from the line and when the speed was finally fixed I’d be getting the most from it. Moreover, the next day James called me back unprompted to check how the line was looking from my end and ask if I’d had a chance to plug into the master socket.

At the moment I am reading “Managing Humans” by Michael Lopp (aka Rands). In one chapter on outsourcing he describes an interaction with a customer support person for his broadband provider and talks about all the “idiot filters” he has to go through just to be able to get to talk to a human being. I can clearly say that A&A have no idiot filters in place - they assume that you know what you’re talking about and I doubt they’ve ever told anyone to re-install their operating system.

There are other things which set A&A apart, such as the card in a clear plastic envelope attached to the top of the router with all the IP addresses on one side and, on the other side, all the login details for the router and the ADSL line so that you can fully access the router or, if you prefer, set up your own. All of this is the opposite of most service providers who try to lock you down and prevent you from accessing anything, presumable on the basis that they think you’ll screw it up. I can, without hesitation, recommend A&A if you are a techie looking for broadband in the UK. Oh, and there’s no 12-month lock-in, just a modest (£5) disconnect fee, or no charge if you are migrating away to another supplier. Oh (2) they provide static IP addresses as standard.

The other company, which I have been dealing with for a while, is FogCreek, makers of FogBugz, a very fine commercial bug tracking system. Why would you buy a bug tracking system when you could just use Bugzilla? Well, it’s like Apple equipment. For the most part it just works. There are lots of little details in the user interface which show the care and thought which have gone into it. It’s clear that the FogCreek guys use it every day as part of their environment which means they have covered most of the things you would want from a bug system and some things I certainly never considered. But also there’s the fact that these guys are so nice to deal with that you almost want to give them your money anyway. I recently moved from hosting my own instance of FogBugz to using the On Demand service. I had, however, recently renewed my support contract on my locally hosted licenses. I mentioned this to support. “No problem”, they said, “we can either send you are refund or apply the balance to the On Demand service.” No hesitation, no quibbles about contract, nothing. Just treating me like a human being and like they care. Many of you may know that FogCreek is Joel Spolsky’s company and already read his blog or have read his books. Joel definitely lives by his no-nonsense philosophy which basically says if you treat your customer as king and treat them fairly then you don’t need to worry about giving refunds or such as these will be vastly outweighed by the revenue generated by repeat business and recommendations generated by the goodwill you’ve created.

These companies demonstrate how I expect to be treated as a customer. Not only that, but their attitude and approach reflect a high degree of confidence that the service they are providing means that customers won’t want to leave so they don’t have to have financial penalties to stop you going. I just wish all companies were like these.

I’m a FogBugz user. I think it’s a great bug-tracking system. So much so that I paid money for it rather than opting for BugZilla. It looks like it’s about to get even better with the new version 6 release.

One of the key new features is Evidence-Based Scheduling. Joel has posted a long article describing it here. What really cracks me up, though, is that the article is really just describing the planning and tracking method used in Extreme Programming (that XP, not the other XP) and other Agile methods.. The foundation of estimation in XP is basically “estimate that you will do as many units of work next week as you did this week”, and by getting the developers to estimate the tasks and then track how much time they took to actually deliver them, you get a very good idea if you are ahead or behind the estimates.

Even this description of what to do if the schedule says you can’t deliver everything by the desired date:

4) A schedule is a box of wood blocks. If you have a bunch of wood blocks, and you can’t fit them into a box, you have two choices: get a bigger box, or remove some blocks. If you wanted to ship in six months, but you have twelve months on the schedule, you are either going to have to delay shipping, or find some features to delete. You just can’t shrink the blocks, and if you pretend you can, then you are merely depriving yourself of a useful opportunity to actually see into the future by lying to yourself about what you see there.

is pretty much exactly the same as what XP describes, though XPers do it by laying index cards out on a table, rather than putting blocks in a box. What is most striking about the article is the way it talks about putting so much of the responsibility and control into the hands of the developers, as opposed the traditional method of “project planning by dictum” controlled by the management or the client which is what leads most projects to fail. This idea of sharing the responsibility and, more importantly, the authority, is a cornerstone of XP and Agile. Trusting the developers to make the best estimates (because, after all, they’re the ones who actually do the work) and not letting managers and the client override them are key. In XP parlance, “time, resource, scope - the client can have control over any two of those, as long as the team is given control of the third one”.

Joel claims not to be a fan of Agile techniques, yet time after time, he has published articles describing the way he develops software which are simply Agile methods re-cast into his own metaphors. For example, in his article on the Project Aardvark Spec he makes the claim “I can’t tell you how strongly I believe in Big Design Up Front, which the proponents of Extreme Programming consider anathema”. However, the spec document that he links to at the end of the article would be completely unrecognisable to anyone who has worked with true Big Up Front Design in the past, and contains nothing, apart from the coding conventions which he admits doesn’t belong there in the same article, which you wouldn’t find written on an index card in an Extreme Programming shop.

Come on Joel, it’s time for you to finally admit that you are as Agile as they come, even if you’ve expressed all the ideas and techniques in your own way (which is a totally acceptable thing to do.) If you don’t think so then please read the Agile Manifesto and tell us which bits you disagree with.

A post by Fake Steve Jobs led me to Jason D. O’Grady over at ZDNet who has blogged about how much he hates the new Apple keyboards. I bought one when they came out to try it out and since then I have replaced they keyboards on all my Macs with the new ones.

It seems that the feel and design is something you either love or hate. I really love the low-profile light touch on the keys. I am a programmer and I use my keyboard probably 10 hours a day. Whenever I use an old-style keyboard now all I can think is how clunky and heavy it feels. It feels like I’m using an original IBM PC keyboard. Interestingly, my Mac Book Pro keyboard, which always used to feel light, now feels quite solid and I actually enjoy using it more now that I’m using the new desktop keyboards.

I don’t mind that Jason doesn’t like them. It really comes down to a matte of taste. His other objections centre around the design and here I think he is on shakier ground. I love the low-throw of the keys. They give enough click sound to give me the feedback I need without the tactile feedback. And objections to the change in the location of they various function keys are really just complaints about change. Millions of new Mac users in the coming year will never care about that and it’s something that most people can easily adapt to.

But most importantly, the keyboard and mouse are two of the most easily replaceable components on a computer. Okay, there may not be as many Mac-specific keyboards on the market as there are for PCs, but there are still plenty to choose from, so if you don’t like the one that Apple puts in the box then go get another one. Personally, although I love the design of the Mighty Mouse, I don’t really like using it so I use a Microsoft Intellimouse (actually, I’ve always felt that MS were a far better at hardware than software) which, of course, works fine.

I agree that it would be nice to see a full-sized bluetooth keyboard, but I’m still looking forward to receiving the cut-down one when it finally ships to me here in the UK.

My point in all of this is that I’m fine with Jason hating the new keyboard, but please don’t try and justify it with opinions dressed up as objective design complaints. There’s far too much of that already in the world.

Well, they’ve finally reaped the full value of their actions. SCO have filed for Chapter 11 bankruptcy protection 5 years after making claims that Linux violated their patents because of code supplied by IBM. They never produced any evidence in court of the violations and carried out a cynical campaign to try and extract money from companies using Linux with the promise of not suing them if their claims were upheld. This was pure protection racketeering and only a few paid up - not even enough to cover their massive legal bills.

Many people seem unaware that this SCO has nothing to do with the Santa Cruise Operation that once contributed hugely to the development of Unix back in the 80’s. That company was bought out a long time ago and the purchaser changed their name to SCO to try and reap some of the positive image, but they’ve never really been anything other than a destroyer of value and never done anything except threaten people with lawsuits.

The final blow was when a court ruled that, although they bought UNIX from Novell in 1993, somehow they did not buy the copyrights along with it (leaving open the question of what they did actually buy.) A company of lawyers, and not very good ones at that.

RIP SCO.