David Richards WANdisco Blog.

What a Git!

noreply@blogger.com (David Richards) — Thu, 17 Jul 2008 22:47:00 +0000

Linus Torvalds put the cat-among-the-pigeons after his rather clumsy attack on Subversion, CVS, Perforce and anything else that was specifically not a disconnected repository in a presentation to Google employees last year. Linus is of course obsessed with the Linux project to the extent that he thinks every company developing software should operate like the Linux project. His comments included:

Subversion has been the most pointless project ever started.

Subversion used to say CVS done right: with that slogan there is nowhere you can go. There is no way to do CVS right.

If you like using CVS, you should be in some kind of mental institution or somewhere else.

Bitkeeper is the only commercially available distributed SCM solution.

The trouble is of course that not all commercial companies operate like the Linux project, in fact very few do. The idea that software developers operate from within a remote cave, rather like a hermit in the middle ages, is more than a little far-fetched.

Git, BitKeeper, Mercurial and Bazaar are collectively referred to as ‘Distributed Version Control Systems’ or DVCS for short. But are they really distributed? In a sense, yes they are. The distribution comes from the fact that they work offline and everyone can have a copy of the entire source tree on their local machine. This is a really cool feature if you happen to be working in a loose-knit environment where working in a prolonged disconnected mode is a positive advantage, for example exploring multiple implementations without disturbing the master repository.

The problem is that, at any one moment in time, there is no ‘golden-copy’ of the source code assets, except via unenforceable convention. This presents quite a large problem to most software companies who undertake continuous builds. Working in a prolonged disconnected mode is fine but what happens if your laptop is stolen at the airport? Not only have you lost your work but you also run the risk of someone having access to the entire source code repository.

This pretty much means that companies, much to Linus’s disgust, must adopt some sort of centralized source code repository or do they?

So I’m pretty biased. Actually, no, I’m very biased but I think WANdisco really is a distributed source code management system (for Subversion and CVS) I don’t think that having a centralized repository with distributed developers solves the problems of distribution either.

In the case of WANdisco every replica is a golden copy of the source code repository. The assumption WANdisco makes is that development teams are working collaboratively even across global teams, with continuous integration of their efforts. Here you get the best-of-both-worlds: performance of a local repository for the entire global team, with the manageability and continuous integration associated with central repositories.

Bottom line: Git, Bitkeeper, Bazaar and Mercurial should not be referred to a Distributed Version Control Systems – they are Disconnected Repository Systems and Subversion / CVS users do not belong in ‘some sort of mental institution’ but I would recommend that Linus get’s out of his virtual reality bubble before it completely seals.

Atlassian Partnership - JIRA Clustering / MultiSite

noreply@blogger.com (David Richards) — Wed, 05 Dec 2007 00:53:00 +0000

We announced a partnership with one of my favorite company’s last week - Atlassian. Founded by a couple of twenty-something-year-old Australians they are taking the developer world by storm. The interesting thing is they are doing it one customer at a time and the products retail for about $5K each. With over 9,000 customers they are doing well, very well indeed.

One of the reasons I am so enamored with them is because they have just got this space nailed. They understand exactly how to position, market and sell into the software development business. JIRA and confluence are their two main products and they are selling like proverbial hot-cakes. The number of our own customers who told us that they either were on or wanted to move to Subversion with JIRA and Confluence is amazing.

Like WANdisco, Atlassian is not venture funded. They are a company built from the ground-up where necessity is the mother of invention. JIRA was built, for example, because the company needed a decent defect tracking system. In fact the story is the same across most of their product suite. Not being venture funded is often a critical success factor for an early stage software company. Taking venture money too early can create an artificial marketplace, whereas building products to put food on the table makes you do stuff better than anyone else. Too many early-stage venture backed company's do too many unnatural things like hiring a huge operational infrastructure for a handful of customers - anyway don't get me started on that topic or we could be here all day.

Back to blatantly marketing our new stuff... We produced 2 great new products: JIRA Clustering and JIRA MultiSite. With JIRA MultiSite we have transformed JIRA into a distributed server implementation thus eliminating WAN latency and, by default, creating a series of globally distributed failover nodes. JIRA clustering, as you could guess, is a clustered version of JIRA that facilitates massive scaling.

From a selfish standpoint, we are not only excited by the partnership but we have also proven that we can embed our secret sauce (DConE) into a database-centric application. Mathematically we always knew it could be done, but there's nothing like seeing it with your own eyes. Secondly, and how we missed this in the past I don't really know, we have what is possibly the best clustering product in the world. Due to our unique architecture we scale disk, memory and CPU. There is no single point of failure, as you may see with traditional clustering architectures that rely on some sort of cache shield, we have a shared-nothing architecture. We will be announcing 2 other clustering products in the very near future...

Technology from Death Valley to Home Depot

noreply@blogger.com (David Richards) — Fri, 30 Nov 2007 07:50:00 +0000

After much consternation, my wife actually managed to persuade me to go on an RV trip over the Thanksgiving break. Head-shaking we set off on a 1,100 mile 5 day trip from our home in the East Bay to Sequoia National Park, Death Valle y and Yosemite National Park.

Being a Brit, 1,100 miles still seems like from here to eternity. In fact I remember when I first arrived in the US, about 11 years ago, I asked for directions and was told it was "just up the road". Now where I come from "just up the road means" a quick stroll. So I set off walking until I discovered "just up the road" in the US can mean almost 100 miles.

So, after driving "just up the road" (453 miles) we arrived at Panamint Springs, Death Valley. To my amazement not only did they have just about the best selection of beer I've seen on this side of the Atlantic; but they also had a high speed wireless Internet connection. Maybe I'm naïve, and maybe I shouldn't be surprised about these things, but I am. It's wonderful to see how technology can change everything. Unfortunately, after connecting to the Internet, it became apparent that the English soccer team had (yet again!) failed to qualify for a major tournament, and suddenly remoteness and the cold Stella seemed a much better proposition.

Technology can and should be beneficial and make positive change. I really couldn't imagine my life without GPS navigation - before GPS Mr Magoo would have got from A-to-B better than me. As an early adopter of VOIP, I can barely remember what it was like to pay for calls to Europe. On the flip side of this there is implementing technology for the sake of it.

For example, some complete &*@#%! stole my notebook from my car. When I discovered what had happened I called the local police, who arrived in less than 10 minutes. The very efficient policeman took copious notes and told me to contact his office for the official police report. I waited a few days, called the station and was told the report was written, but was "somewhere in a workflow approval system". Being an old EAI vendor I could empathize with her and she confided that "before this they'd just type them up, put them in a file and they'd be done in a day or so". Many of the EAI vendors over -promise and under-deliver and often, as in the case of the local police station, make things worse than they were. So much so, the industry consortium had to remove the words EAI from its' name.

Those self check-out machines at Home Depot are another example. I was reading the other day how a guy in Seattle took a crowbar to one of these incredibly frustrating machines after he accidentally hit the Spanish button. I can sort of understand this - and he did leave the store without completing the purchase of the crowbar...

Give me a KISS!

noreply@blogger.com (David Richards) — Fri, 18 May 2007 21:24:00 +0000

One of the things I feel passionately about is ease of use. I’m one of those guys that doesn’t normally read the documentation – the theory being that you should be able to intuitively use any product. Take Google’s Adwords product for example. I really hate to compare anything to Google by the way – everyone in the technology industry wants to compare just about anything to Google – then talk themselves out of doing anything because “Google are already doing that”. What a load of tripe – if that was the case nobody would ever start a company. Anyway, I digress. Google’s Adwords product was just about the easiest product I’ve ever used. When Adwords was first launched we managed to get an ad up in a couple of minutes that served hundreds of hits to our website. It was intuitive, clean and simple.

Pretty much everyone agrees that Google changed the game in online ads. Yet there was nothing innovative about online advertising. Yahoo and Microsoft had been doing it for years. Google though made it easy. I didn’t need to read a manual, I didn’t need customer support - quite simply it was the best product on the market.

Compare that to Overture. The interface was confusing, my anti-spyware blocked cookies that the product needed to log me in, it took days not minutes to get an ad up.

Speaking as a software vendor, we all need to spend more time on making this stuff easier. At WANdisco, we’re big boys and girls about this stuff and admit to ourselves where our product could and should be improved. Ease-of-deployment and ease-of –use are top of the agenda right now. Our product solves a terrifically complex problem; but we don’t want our users to see or deal with any of that complexity. It is a trap to assume that users have special skills – sure some do but they are usually early adopters. Einstein's maxim that "everything should be made as simple as possible, but no simpler" probably says it all.

Subversion Comes of Age

noreply@blogger.com (David Richards) — Fri, 20 Apr 2007 01:04:00 +0000

Over the past few weeks I’ve had the great pleasure of meeting quite a few senior mangers and executives from some of the world’s largest companies to talk about source code management (SCM). The recurring theme from them all has been their readiness to adopt an open source product to manage their source code.

The product in question is Subversion.

Now that’s pretty cool, but now consider that most of these companies are ripping out proprietary technology in favor of Subversion and you have an industry trend that may already be upon us.

This isn’t new though – who would have thought that Linux would challenge Windows, JBoss would challenge BEA or even that MySQL would dare to take on Oracle. By the time the market realizes that open source is a legitimate challenger, it’s already happened. The result is nearly always a commodity market with terrific downward price pressures – good for customers, bad for vendors.

Subversion is showing those characteristics. Managers seem to understand that the SCM repository is commodity. They should not be paying millions of dollars in support and maintenance for that. Add administration costs to the equation, and proprietary SCM is an expensive proposition – even for companies with huge IT budgets.

So is cost the reason so many are looking to Subversion? Well, sure its free, but that’s not enough - CVS is free remember, and growth for that product has slowed dramatically. Subversion is really liked by the development community so there’s lots of innovation. Many of the annoying things with CVS such as a lack of atomic commits are fixed with Subversion. This makes Subversion a whole lot easier to adopt from a large enterprise perspective. What’s more, the Apache/BSD license is not as intrusive as GPL, particularly for vendors looking to OEM.

So the future is looking good for Subversion, which is great news for our company, WANdisco. In fact the timing could hardly be better. Many companies looking at Subversion usually have multiple sites so they go and look for a Subversion Multi-Site product. A cursory glance at Google will quickly get you to our active/active replication solution for Subversion. Disaster recovery is also a significant issue, and we can provide that on a WAN scale… more about that next time.