Prediction #8: Eclipse and Visual Studio dominate the IDE market, become even more boring

by Mik Kersten, January 31st, 2011

I miss the heyday of the IDE wars. Seeing David Intersimone’s demo of JBuilder at OOPSLA 1998 opened my eyes to the way that tools can multiply our productivity. The competition that ensued between Borland, Microsoft and others brought us code assist, refactoring, structured editing, GUI builders, and a far more integrated developer workbench than the windowing system and command line could provide. In 2001, the landscape changed dramatically when IBM killed a significant part of the developer tools business by open sourcing Eclipse (later re-entering the tools business by acquiring Rational). The open sourcing of Eclipse was a success and created an unprecedented platform for tool innovation. But Java developers started expecting their IDE to be free, and the IDE business was gone for all but the most inspired tool builders, such as JetBrains.

The base IDE feature set has now been standardized and commoditized. In other words, it has become a lot more boring. To get buzz on Twitter, vendors have to pull zany moves like running a thin subset of IDE functionality in a web browser. The bleeding edge aside, we won’t see any dramatic departures from the base IDE feature set anytime soon. However, we’ll continue getting many nice tweaks and features, as you would expect with mature tools like cars or can openers. As with other commodities that need to differentiate their extremely similar feature sets via slick dashboards and sophisticated cup holders, we will see some welcome improvements on the fronts of eye candy and usability. This kind of tweaking is already surfacing in Visual Studio 2010, which saw an investment in user experience and graphic design refresh that put it ahead of the curve.

While the base feature set of code editing, building, debugging and deployment solidified, tool innovation will now occur on top of the commoditized IDE platform. The major shift in the IDE over the last decade has been to shift from end-user tool to tool platform. And like any software platform, the IDE now needs to evolve more slowly because the innovation is happening on top of it. Beneath the IDE we’re swapping out the deployment model to be cloud-based, as with the SpringSource Tool Suite’s (STS) direct cloud deployment. Atop the IDE we are layering on all sorts of collaboration features, recognizing that developers spend as much time collaborating around code as they do hacking it. I can no longer imagine working without having rich and IDE-based access to a task board and release plan.

The IDE has multiplied productivity by integrating, then personalizing developer’s workbench

The maturity of the Eclipse 3.6 and Visual Studio 2010 platforms has made it possible to bring the developer’s entire coding and collaboration activity into a single workbench built on plug-in APIs and shared data models. In 2011, we will see the remaining APIs completed, such as Mylyn’s new APIs for IDE-integrated code reviews, distributed version control, continuous integration and documentation. Whereas last decade’s IDE was all about integration and helping developers stay in the flow of their work, the next decade of the IDE will be about personalization. We are shifting from the IDE being a view on the entire system to the IDE being a view of the parts of the system and planning that are relevant to the developer. The shift might sound less exciting, but its impact on developer productivity will be just as pronounced, and is necessitated by the steady growth of software complexity. The integration of Agile, ALM and a task-focused workflow will drive the IDE’s shift from an integrated workbench to a highly personalized one, incorporating the focus and flow needed to keep developers working on extremely complex software happy and productive.

Boeing’s 787 Dreamliner is reported to run 6.5 million lines of code, approximately the same number as the core Eclipse platform. The whole of the 2010 Eclipse Helios release train was 33 million lines of code. Add to this all of the third party extensions, and today’s IDE platform ranks with some of the most complex artefacts produced by humanity. Due to the scale of investment, marketing and ecosystem management, as with commercial aerospace, we will see the two big players dominate the IDE market. The small and niche players will continue to exist, ranging from the boutique IntelliJ to JDeveloper for database development. But for those rolling out an IDE across their organization in 2011, the number and availability of third party extensions will be a more important decision factor than the base IDE feature set. ISVs and cloud/PaaS vendors will increasingly use the IDE as an on-ramp for their offerings, with Eclipse and Visual Studio as the clear targets for their integrations. Following the user-centric success of the Embraer E-190, which provides a superior cabin experience to similar offerings by the two aerospace giants, a renewed focus on developer productivity and Agile collaboration needs to be the driver of change in the IDE over the coming years.

6 Responses to “Prediction #8: Eclipse and Visual Studio dominate the IDE market, become even more boring”

  1. Prediction #8: Eclipse and Visual Studio dominate the IDE market, become even more boring Says:

    […] The competition that ensued between Borland, Microsoft and others brought us code assist,… [full post] Mik Kersten Tasktop Blog eclipsemik on eclipsemylynopen source 0 0 […]

  2. Jason O'Keefe Says:

    I agree that the next wave of IDE success will be in the most agile, flexible and collaborative tools around. The platforms themselves will likely move into quiet ubiquity as opposed to being “sexy”. MyEclipse (disclosure: I work for Genuitec) began to address this about a year and a half ago for Java developers by creating the most flexible and collaborative-driven Eclipse-based IDE. Here’s hoping the “wars” are over, but the prevalence and evolution of good tools is still in its infancy!

  3. Mik Kersten Says:

    Jason, while I agree that the platforms are headed into quieter ubiquity, I do wish we would do more on the Eclipse side in terms of the slick look and feel. It bugs me that Visual Studio 2010 now looks more slick than Eclipse due to the extra investment in a refresh of the look-and-feel. Tasktop did a considerable amount of work with Microsoft in order to make Eclipse light up on Windows 7 but there is still work remaining in modernizing some of the widgets and shell, such as: bug 325795: support Windows Vista and 7 Aero Glass shells.

  4. Olaf Bergner Says:

    What I miss in your IMHO otherwise very accurate assessment of today’s IDEs’ future directions is adressing the most pressing problem I face in my everyday life as a working developer: scalability.

    Frankly, Eclipse & Company suck at handling the kind of code bases I regularly work on, i.e. having a LOC count in excess of 10 million. I have to disable all the bells and whistles – automatic compilation is most prominent amongst those – just to get Eclipse going. Don’t even dream of large-scale refactorings across the whole code base.

    All those niceties to come will surely add to today’s IDE’s perceived polish and are surely welcome. And yet I refuse to call a technology mature that regularly fails to deliver its promised productivity gains as soon as a project gains considerable size and complexity.

    But addressing those problems would probably turn out to be hard work, it might even require redesigning core components. Given the currently prevailing perception that the age of innovation in the IDE space is largely over I am afraid I won’t see my hopes realized any time soon.

  5. Mik Kersten Says:

    You bring up a very good point Olaf. Given reduced investment in core IDE features and faster growth of system size than desktop/laptop performance, how will developers be productive when working with 10M+ line systems?

    In terms of raw IDE build performance for Java systems, I think that we will only see incremental improvements for the foreseable future. The 3.4 release Eclipse added multi-threaded compilation, and thanks to proper incremental compilation can happily build a multi-million line systems on my laptop’s Core i5 processor. Eclipse is notably slower for compilation of some other languages, and more complicated build processes can also cause slowness, such as those involving heavy code generation. Given ongoing system growth towards 100M lines systems, we are approaching a losing battle in expecting the IDE to build the entire system. Instead, I think that we need to do more to move towards a the world where the only authorotative build is in the CI system and developers only attach their IDE to the parts of the build that they need to work on. This has been a long-standing practice on and on the Mylyn project in particular, and what we need now is the next round of tooling automation needed to make that way of working easier to deploy at the organizations who still have developers suffering from a sluggish IDE.

  6. Tasktop Blog » Blog Archive » Newsletter: New SmartBear CodeCollaborator Connector, Visual Studio public beta now available Says:

    […] Prediction #8: Eclipse and Visual Studio dominate the IDE market, become even more boring […]

Leave a Reply