News Blog

At lunch with Michael Coté from RedMonk on Wednesday, we talked a lot about how open source has really split into "free" and "open source," with the former typically associated with basement developers and Apache licenses, and the latter generally associated with the General Public License and some set of enhanced features.

As I was following Coté's Twitter feedearlier, I started to wonder whether everything really will go to the cloud and all of our open-source musing will go away, as software becomes consumed versus installed.

Realistically, there is a vast array of software that really can't move outside the enterprise in the foreseeable future. Consider, for example, banking and stock-trading systems, or telecommunications infrastructure. On the other hand, consider pretty much everything else. Even when you take into account the complexities of back-office systems, odds are that in a green-field situation, you could find a software-as-a-service application to solve your problems.

So here's the paradox that I think about: Let's consider a company like Google, which writes, buys, and installs a lot of software. Some is unique to its business and isn't available as an online service. Other products are packaged applications. Yet it wants the rest of the world to stop buying software, instead just consuming it from Google.

I'm not seeing a way that on-premise software disappears forever...

Following on its successful launch of Red Hat Enterprise Linux as an Amazon Elastic Compute Cloud service, Red Hat is now offering the JBoss Application Server on EC2.

It's yet another example of open source truly becoming a Web-enabled service, rather than a mass of packaged bits and bytes. And it comes at a reasonable price:

Red Hat is charging a fixed subscription rate of $119 per month for JBoss Enterprise Application Platform, or a variable fee, starting at $1.21 per instance, per hour, with fees depending on the size, bandwidth, and storage of the services purchased...Customers can either license JBoss on EC2 from Amazon and receive a virtual image of the software, or make their own subscription of JBoss available on Amazon's compute cloud.

The more Red Hat and others can deliver their software as Web services, the less trouble there will be with getting a fair return on R&D investment in commercial open source. It makes a development service into a Web service, which looks an awful lot like a product that people are used to buying. Maybe an answer to Savio's fair critique?

Google Docs, the online office suite from the search giant, now has some limited but still useful support for PDF files.

PDF files now show in Google Docs' interface.

People using the service now can upload and view documents encoded with the widely used and now standardized Portable Document Format initially created by Adobe Systems. People also can transfer PDFs stored on the Web. (Look below for a screenshot showing the two-pane PDF view.)

The move, announced on the Google Docs blog Friday, isn't much of a surprise. In addition to the fact that it makes eminent sense, close observers already had begun seeing signs that hinted at imminent Google Docs PDF support.

Google Docs, still in beta testing, competes with Microsoft Office but is relatively primitive when it comes to feature support.

However, because it's Web-based, Google can add new features relatively easily; users simply use the Web site, and they appear, one of the chief advantages of the software-as-a-service approach. And given that Google's three big areas of focus are search, ads, and applications, expect lots of resources to be poured into this area.

I found the PDF support snappy and very handy. However, my quick test of the service showed some rough spots with the PDF support.

For example, I couldn't find a way to zoom in or out, which definitely is essential, even on ordinary 1024x768-pixel screens. Being able to hide the minidocument page view pane on the right, which lets you scroll quickly through the document, might help.

Search also doesn't scour the contents of PDF files, a feature whose significance Google, of all companies, presumably understands.

Editing has a long way to go. You can't type text in a PDF, though you can export other Google Docs files to PDF. And copying uses a peculiar box to select text, not the familiar cursor with highlighted words.

You'd better have a screen at least 1024 pixels wide. Most of us with PCs these days do, of course, but what about support for mobile devices?

I also didn't like one user interface moment: the site offered a very unhelpful error page when I tried to upload a file exceeding the 10MB size limit.

Overall, though this is a big step in the right direction.

An example of Google Docs showing a PDF file.

Sun Microsystems is launching a program to help software makers convert their existing, on-premise applications into software-as-a-service offerings.

The Solaris On Demand program, announced on Wednesday, is targeted at independent software vendors. Sun offers the software, hosting and services to convert applications. Sun says it is partnering with NaviSite. AT&T's USi Communications, and NTT Europe to provide hosting services.

Sun, along with partners, will offer a 90-day proof-of-concept trial to give independent software vendors access to hosting, hardware managed services, and backup services. The company guarantees a service level agreement of 99.5 percent uptime.

The service price varies by ISV, based on the scope of work needed, according to Sun. The price "can vary from a thousand dollars to hundreds of thousands of dollars based on how many seats are required," according to Vince Vasquez, a business development manager at Sun.

Here is the first third of CIOZone.com's list of the top 60 fastest-growing, public software companies with revenue of at least $150 million.

(Credit: CIOZone.com)

CIOZone.com has ranked the top-60 fastest-growing (public) software companies of at least $150 million in revenue, with VMware leading the pack and Red Hat claiming 12th place with a 33.6 percent growth rate. Not bad for a company that gives away its software for free.

But then, perhaps it's not surprising since Google, ranked second on the list, largely does the same.

What's most impressive in the list, however, is the growth rate being sustained by Oracle and Microsoft, because they're growing from a much larger base. It's fantastic that Red Hat is growing at 33 percent on a ~$500 million base. But Microsoft is growing by 25.7 percent on a $57 billion base, and Oracle is rising 24.9 percent on a $20 billion base.

That's amazing.

For smaller companies (more than $50 million in sales and less than $150 million), Omniture (Utah-based - hurray!) leads the pack with a 79.5 percent growth rate on a $143 million base. Not too shabby.

While the list is predominately comprised of proprietary-software companies, CIOZone points out that these vendors are succeeding precisely because they, too, are changing the game from a proprietary license model to a subscription model:

Software that increases the efficiency of corporate data centers, or that runs as an Internet-delivered service, is on a roll. Most of the fastest-growing software companies are doing one or the other....

While it was a good year for many software vendors, it was particularly good for companies that have discarded the older paradigm of install-and-maintain and are making their products available as online subscriptions.

This bodes well for open-source companies, for two reasons. One, many SaaS companies depend on open source (and will pay for it) to run their businesses.especially as an increasing number operate with a dual-mode model wherein they make their software available as an open-source download, with a pay-for-SaaS model to complement that. Kaltura, Loopfuse, and others are adopting this model, and I think it promises to be a highly profitable model for those that can pull it off.

Gartner seems to have checked its former open-source blindness at the door, and is now suggesting that 90 percent of all software-as-a-service providers will adopt open source in their infrastructure by 2010.

Cost cutting will lead to the move, said (Gartner)...Open source will be used in the operating system, application server, and at a database level and will make up 30 percent of an application.

Of course it will. Open source is the foundation of software innovation in the 21st century.

One big question remains, however: will open source provide SaaS' free lunch or will there be a quid pro quo? With AGPL and the Open Software License, it will be the latter. With 20th century open-source licenses, however, SaaS gets a free ride.

The more open-source software we want, the more open-source software we should be prepared to pay for, whether in cash or contributions. This is a fair exchange, and will provide the basis for a robust SaaS software economy for many, many years to come.

Lots of large system and software vendors, even (or perhaps especially) those who have traditionally focused largely on enterprise sales, talk loudly and often about the midmarket these days.

Or they use terms like "SMB" (small and medium business) or "SME" (small and medium enterprise)--categories that aren't really the same thing but nonetheless often get used more or less interchangeably to denote companies with about 100 to 1,000 employees.

The reason for the interest is pretty simple. There are far more smaller businesses than there are larger ones--especially in developing economies. Midmarket IT spending is also growing quickly in many categories. As a result, it's not especially surprising that even those vendors most accustomed to selling to enterprises are itching to boost their midmarket share as well.

The challenge they face is that IT at midmarket companies bears, at best, a passing resemblance to that of enterprises. Development and operations staffs are small and are far more likely to be made up of generalists than specialists. Furthermore, most selling to midmarket companies takes place through regional or vertical market partners of some sort. Thus, the vendor seeking to increase midmarket footprint typically has to put together different types of product packaging, if not entirely different products, and craft go-to-market (GTM) approaches that differ in substantial ways from those supporting large enterprise sales.

Given these complexities, it's natural that historically enterprise-y vendors have taken awhile to craft successful midmarket plans. Even IBM, which has long had a strong midmarket presence with System i (and the AS/400 that preceded it), has gone through numerous GTM and organizational iterations to improve its ability to tap the midmarket with other product lines. However, most large vendors have made progress. The best are doing quite well (as we've noted in the case of Hewlett-Packard for example). Others, such as Sun Microsystems, have at least made improvements to their partner programs even if their overall grade remains middling.

Expanding market reach in this way is all well and good. Certainly, many small and midsized businesses continue to run their IT departments in much the same way as in the past on in-house systems running a combination of packaged and custom applications.

However, we're also struck by how tenuous the connections are between many vendors' midmarket planning and any sort of software-as-a-service (SaaS) or broader cloud computing initiatives.

Computing isn't going to jump from the data center to the network overnight; if nothing else, the rate and specifics of such a shift are matters for spirited debate. But as a general direction for computing, it's hard to make much of an argument.

It's also hard to argue with a contention that, again as a general rule, SaaS should have more near-term appeal in the midmarket than in the larger enterprise. The list of reasons why is substantial: midmarket companies have smaller IT staffs; they tend to use more pre-packaged and less complex applications; their IT infrastructures tend to have fewer "moving parts" (and therefore better lend themselves to carving out pieces to run over the network); and existing relationships with midmarket ISVs and VARs could make the transition to SaaS applications from those partners fairly natural.

Or, to put it more bluntly, enterprises may find running their own IT infrastructures costly, but for many midmarket companies, it's genuinely hard and even harder to bring new applications and the like online.

Therefore, the disconnect at many vendors between forward-looking SaaS and cloud computing strategies and their more tactical midmarket plays is, at the very least, a strategic oversight.

Metrics tend to play a middling role after a social-media company is established. With a few notable exceptions, most sites hit a peak then flatten out. The same can be said for open-source projects. There are peaks, but really what you want is consistency in the numbers. Hyper-growth is not sustainable and you will eventually saturate the addressable market.

One of the things open-source projects tout is their number of downloads. Download metrics are often very flawed but can help tell the story of how big the market potential is. Social-media companies talk about their number of registered users. Neither talks much about retained users unless the metrics are very impressive.

Viral models

(Credit: Andy Chen)

What does running software in the network mean exactly?

This is one of the questions that users are exploring as they start to increasingly poke at what "cloud computing" means for them.

On the one hand, cloud computing can refer specifically to running some sort of fixed software service--frequently through a browser's user interface--over the network. This is cloud computing in the Web 2.0 sense. We don't necessarily even think of Flickr, Facebook, or Google as "applications" as such. At the other end of the scale, services such as Amazon's EC2 and S3 just rent bare CPU cycles and storage capacity for whatever software a user wants to load up.

However, between these extremes lies a continuum of customization and malleability. Application programming interfaces that allow third-party customization and extension are rapidly becoming a de rigueur companion to software as a service. At the same time, virtual appliances and other predefined software loads offer at least a degree of preassembly when renting raw computing by the hour.

Tuesday's announcement by Mosso, a start-up funded by hosting provider Rackspace, offers up yet another variant. The core concept behind Mosso's Hosting Cloud is that many Web-based applications or sites are built up using largely common stacks of technologies such as PHP and MySQL databases. Mosso takes advantage of this fact by providing the means to provision applications running on one of these common stacks. Mosso is effectively offering cloud computing at a level of abstraction more akin to that of a Web hosting provider. For example, Mosso takes care of patching and updating the operating system and other software stack components. This is unsurprising given Rackspace's historical business, but it's a bit different than what's generally discussed in the context of cloud computing.

A user sets up a site by logging into Mosso's management application, entering a domain name, the technology stack to be used (Mosso supports Windows/.Net as well as Linux), and additional services required--such as databases. Mosso will then provision the site on a cluster of servers at which time the user can upload custom code.

The big difference from a typical hosted Web site is that Mosso monitors the site's resource use and will scale up available hardware resources as needed automatically. The pricing model is as follows:

Base pricing is $100/month, which includes:

• 24x7 live technical support (phone and chat)
• 50GB disk space
• 500GB bandwidth
• 3 million Web requests/month  (A Web request is the retrieval of any item from a Web server, i.e. a Web page with two photos counts as three requests)

Additional disk space is 50 cents per gigabyte, bandwidth is 25 cents per gigabyte, and requests are 3 cents per 1,000 requests.

Mosso does not currently provide any means to throttle or otherwise limit the traffic or resource use by a site. This seems reasonable enough in the context of businesses that would typically be more concerned with their site going down than in having an unexpectedly large hosting bill at the end of the month. In addition, by partially pegging charges to Web requests, Mosso is aligning its fees to a measurement that has direct relevance to many companies operating Web sites--especially if they are advertising-supported in some way.

Writing Defining Cloud Computing last month really crystallized for me that it would be a mistake to narrowly define this trend as only about Web 2.0 or software as a service. Announcements such as Mosso further emphasize this point. More and more computing may go out into the network. But the way that it moves into the network will take a multiplicity or forms--especially as users experiment in these early days.