Blogs

RedMonk

Skip to content

Novell’s Hovsepian: Off the Mark

In remarks that can only be labeled ill-considered, Novell’s President and COO Ron Hovsepian pretty directly attacked Sun’s OpenSolaris project. Now before I get the Enderle/Economist treatment from Ashlee ;), let me be up front and say that Novell is not a RedMonk client, while Sun is. So feel free to keep that in mind when you consider the following. Please also keep in mind, however, that I know and have a lot of respect for many of the Novell folks like Miguel (I benefit from the fruits of their labor every day, as I rely on a number of Mono based applications).

Hovsepian’s made no shortage of bold claims recently, as others have noted. Perhaps the most notable is his claim that Linux on the desktop is set to take off within 12 to 18 months. That kind of commentary, however, is to be expected from a vendor that’s really focusing on that area for expansion. And for what it’s worth, I’ve heard much the same message from others with substantial reach into enterprise and government. Novell’s been responsible for a lot of innovation (Xgl, etc) in the desktop arena of late, and should take all due credit for that. [1]

But in discussing OpenSolaris, Hovsepian – assuming that his words were not taken grossly out of context – says some truly bizarre things.

  • Contention #1: Open Solaris is “marketing open source, rather than a true technical open source”

    While I’m not even sure what that really means, I suspect he’s not grumbling about the usual suspects – the non-GPL CDDL license that OpenSolaris uses, or the fact that some of the tools necessary to build OpenSolaris weren’t present at the initial launch. Instead, it seems as if he’s trying to characterize the OpenSolaris project as window dressing, something Sun did solely for the PR bump of having an open source operating system.

    Given the costs associated with open sourcing software, this argument is not particularly supportable – the legal fees involved with open sourcing a piece of previously closed software alone would outweigh any potential “marketing” value. But it also belies the real successes that community has had, whether in distro or participation form.

  • Contention #2: “The whole spirit of open source is to have one base of code”

    Maybe some of you from the Linux side of the aisle are inclined to agree with Hovsepian’s criticism of OpenSolaris and its respective community, but this comment is more or less incomprehensible. I’m not even sure where to start, but let’s look at the major application categories. Operating systems? There are several besides Linux (and it’s been argued to me recently that Linux isn’t even a single entity itself), the most popular of which is FreeBSD. How about application servers? Again, there’s a lot of choice, with JBoss and Geronimo leading the pack. Java containers? Apache’s got two to pick from – Jetty and Tomcat. Enterprise databases? Take your pick from Ingres, MySQL, Postgres, and so on. Maybe the web server market? Not so much. While Apache dominates the field here, you still have other options including LightHTTP, WEBrick, etc. Hovsepian’s own company, as others have pointed out, supports two open source codebases – GNOME and KDE – designed to address the same market.

  • Contention #3: “Open-sourcing Solaris – while it’s appreciated that you can see the low-level pieces of code – doesn’t move the overall effort of the open-source community further down the track. It creates a fork, which none of us likes.”

    Maybe this explains the above comment; that the one codebase argument was not intended to imply that there should be a single project per application area, but rather that open source abhors forks. The question then is whether or not that’s true. In general, I suppose it is – forks aren’t everyday occurences in major projects, but as Mambo/Joomla illustrates, they can and do happen. But there are two problems here: one, OpenSolaris isn’t a “fork” any more than OpenSuSE is, and two what constitutes a fork. Are Debian, Gentoo and OpenSUSE forks because they use different and incompatible package management systems? How about if they distribute their own custom, patched versions of the Linux kernel?

    Or perhaps Hovsepian is implying that any open source operating system that’s non-Linux consitutes a fork? In that case, somebody better go tell the FreeBSD guys at Yahoo to switch from their fork to Linux.

The interesting thing about this is that the folks on the OpenSolaris list are generally enthused about it, rather than intimidated. They seem to be viewing it as validation that they’re making progress. And I think there’s something to that; where Linux advocates were once simply dismissive about the prospects for Solaris/OpenSolaris, the public and private commentary about the operating system I had once dismissed has gotten markedly more aggressive of late. Bears watching.

[1] Don’t read that as “Novell only innovates on the desktop.” The SuSE folks have done great kernel work over the years, and Novell/Ximian’s Robert Love literally wrote the book on kernel hacking.

Categories: In the Headlines.

  • http://tieguy.org/blog/ Luis

    I’m guessing what Hovseppian should have said was to compress and rewrite #2 and #3 into ‘the spirit of good engineering is to avoid reinventing the wheel when possible’. Here, Sun fails- they could have released the key components as GPL and ported them to Linux; instead they chose to duplicate tons of code and duplicate an existing community. At least to an outsider, OpenSolaris smells very much like a bad case of NIH.

    [And as I’ve argued elsewhere, Hovseppian/Novell’s support of both GNOME and KDE angers me as a stockholder. Novell should pick one to lead the world with one instead of duplicating effort and being merely good at both. Ideally they should also pick the one that presents a united front with Sun and Red Hat to ISVS.]

  • http://jroller.com/page/jaimec Jaime Cardoso

    Yes, and Novell could have released key components of Xgl as a BSD licence and avoiding duplicating efforts by the BSD community.
    It’s kind of sad to see Novell to enter the RedHat bandwagon of OSS=GPL (ops, but we need a browser to fight IE so, let’s make it OSS=GPL+Mozilla) and Good=us => Bad=everyone else.
    It’s sad to see this discussion comming again. Sad and worthless. It’s over. Reality proved Hovseppian wrong. Before the governance model being presented at the Opensolaris community, I could even have understood a discussion of Opensolaris isn’t really open because the contribution model didn’t existed so, no one outside Sun could influence the direction of the code before it was released but, even that point has been addressed and, even if it didn’t, Novell would be the last company that could bring up that issue (not that I was against the Xgl development methodology but, let’s be consistent, OK?).
    Instead of all the baby cry of “mommy, they are mean to us”, Novell should really be proving to the world that they are right. Clean up glib, make the Linux kernel to scale linearly up to 144 cores or more, prove to the world that Opensolaris really is redundant, then, take your cheap shots at Sun for betting on the wrong horse.
    Speaking of the Xgl, why didn’t Novell simply developed upon the work made with lookinglass? At least to an outsider, Xgl smells very much like a bad case of NIH

  • http://www.webmink.net Simon Phipps

    Nice piece; very restrained, Stephen! I have a long list of questions I’d love to ask Mr Hovsepian , but I’d argue especially with his third contention.

    Since we opened Solaris, we’ve seen the *BSD community take OpenSolaris code and run with it, we’ve seen the Debian community experimenting with new ideas, we’ve seen the Linux kernel community trying to better DTrace. I’d argue that, even ignoring the richness opening Solaris has brought into the OpenSolaris community, it’s also injected code, creativity and competition into the wider open source community of communities and even at less than a year old is doing immeasurable good.

    The fact he can’t see that suggests he’s way out of touch with his own open source people (Sun is a huge GNOME contributor) and with the wider communities. I’d suggest that’s the really worrying sign, at least for Novell’s shareholders.

  • http://duckdown.blogspot.com James

    Liferay doesn’t restrict the folks from eXo from borrowing ideas or even code across portals but Solaris prevents Linux from doing so in my limited understanding. If this is even somewhat true, does he have a point?

  • http://www.redmonk.com/sogrady stephen ogrady

    Luis: i can see the point, i suppose, but i don’t think it’s as simple as saying that Sun should have just donated the good pieces of technology. first, it’s technologically non-trivial, as i’m sure you’re aware, and second, there’s the Solaris installed base to consider. while it’s smaller than it was, thanks to Linux, it’s still substantial. given Sun’s fortunes of late, i think it would have been difficult – to say the least – for them to transition to Linux and not lose an even larger portion of their customer base than they already have.

    i’m also a believer that having Solaris around pushes Linux in a useful way.

    Jaime: good points, but while i know a lot of folks would have agreed w/ you re: Xgl, i’m not entirely convinced. open sourcing later in development is not my favorite methodology, but in this case might have been necessary.

    Simon: agreed on the competition angle, but i’m not sure i align with your last contention: it may be precisely b/c he sees that ;)

    James: that is true (although Jonathan Schwartz did verbally give developers the go ahead to port DTrace to Linux at last summer’s OSCON), but i don’t believe it’s a fault that can be attributed to Linux. Linux, being licensed under the GPL, is not compatible with either MPL derivative (such as the CDDL) or BSD licensed projects. it’s the nature of the license. code from OpenSolaris, for example, has been ported to FreeBSD, but this is not possible with Linux.

    to me this isn’t a fault thing, it’s a regrettable consequence of the differences in open source licensing. but i would reject it as a criticism of OpenSolaris specifically.

  • http://jroller.com/page/jaimec Jaime Cardoso

    Stephen, I too agree that the Xgl methodology was the necessary one but I can’t criticize that method on one side and do the same thing on the other.

  • http://www.frenchguys.com Erik Dasque

    Hovsepian is a glorified sales bot who was put in a position where he oversees the product but no real understaning of it.

    He latches onto some sentences heard here and there and tries to include them into his sales pitch.

    Just like most executives at Novell, he’ll underperform and will get canned. But he’ll get a very large bonus, salary and golden parachute. Meanwhile, Novell will prepare their next layoff.

  • http://www.frenchguys.com Erik Dasque

    Hovsepian is a glorified sales bot who was put in a position where he also oversees the product line but without any real understaning of it.

    He latches onto some sentences heard here and there and tries to include them into his shock the world – sales pitch.

    Just like most executives at Novell, he’ll underperform and will get canned. But he’ll get a very large bonus, salary and golden parachute. Meanwhile, Novell will prepare their next layoff.

  • http://dotnot.org Scott Sanders

    Great post, Stephen. Perhaps you could ask Novell directly why they have 3 groupware solutions, 2 of which are open source? http://dotnot.org/blog/archives/2005/03/10/novell-and-groupware-which-way-did-they-go/

  • Miguel de Icaza

    Jaime Cardoso,

    You are wrong on two counts regarding your Xgl claims.

    Xgl was released under the MIT X11 license, the same license that the rest of the X server is released under. The same applies to compiz, as opposed to the majority of the window managers out there.

    As for lookingglass, it was an interesting research prototype, but it was too large (the vernacular used to describe this I believe is “bloated pig”). Who wants to be running the JVM to display their windows on their desktop, as cool as 3D windows might be?

    The core is not really Xgl. Its compiz. Xgl was merely an X server implementation that had the features requires (I believe 3 server extensions, Composite, Damage and Render plus a texture extension to the GL protocol)

    As those extensions are making its way into the native drivers (NVidia has them now at least) there will be no need for Xgl, all you would need is the window/composite manager (Compiz today, others tomorrow, I believe KDE its doing its own).

    Miguel