Memo to Canonical: Don't Repeat IBM's Mistakes
As Canonical polishes Ubuntu to potentially support Google Android and Moblin v2, I’m both impressed and concerned. On the one hand, Canonical continues to change with the times — and in some cases, stay ahead of the times. On the other hand, Canonical runs the risk of repeating some major software mistakes IBM made in the 1990s. Here’s why.
First, the upbeat perspective: By supporting Google Android and Moblin v2, Canonical effectively positions Ubuntu as a flexible, general purpose operating system that continues to evolve on Netbooks and Mobile Internet Devices (MIDs).
Historic Concerns
Now, my areas of concern. Rewind to 1992. At the time, IBM vowed to make its OS/2 operating system a “a better DOS than DOS and a better Windows than Windows.” At first, IBM’s commitment to making OS/2 run Windows applications looked very smart. For a very brief time, OS/2 really was a better Windows than Windows.
But over time, OS/2’s support of Windows allowed software developers to focus on a single operating system: Windows. It became difficult for software developers to justify the extra time and money to write native OS/2 applications. IBM actually paid some ISVs to write 32-bit OS/2 applications but it was a losing proposition.
The bottom line: By emphasizing Windows support, IBM accidentally de-emphasized native OS/2 support.
Canonical’s Approach
Now, let’s apply the IBM OS/2 example to Canonical and Ubuntu.
As Canonical attempts to promote Ubuntu into businesses, native application support and related services (such as Landscape and Ubuntu One) will become ever more critical. And so far, Canonical earns mixed grades on its ISV (independent software vendor) results.
Canonical attempted to rally ISVs around Ubuntu during LinuxWorld 2008. But so far, I haven’t seen many examples of ISVs promoting applications on Ubuntu.
No doubt, developers and hackers who know how to dive into code can make a range of commercial applications run on Ubuntu. But those applications (example: Oracle) still need to be certified on Ubuntu in order to attract interest from small, midsize and large businesses.
By supporting Google Android and Moblin v2, Ubuntu will certainly gain more buzz. But Canonical needs to make sure ISVs are focused on Ubuntu itself.
Follow WorksWithU via Identi.ca, Twitter and RSS (available now) and our newsletter (coming soon).
Spot on .. i love what ubuntu is trying to achieve but somethings times am feeling its could be biting more than it can chew. trying to maintain ubuntu-server, ubuntu netbook remix, kubuntu xubuntu ubuntu, ubuntu mid, edubuntu etc will take it’s toll soon. i would rather it focuses on a few things and do it well. it can create a base from where the community will just take it from there. For far ubuntu is doing a Job as releasing quality products on all fronts. but for how long?
Bigbrovar: Thanks for your perspectives. Your list of all the Ubuntu offerings got me thinking … How come there’s essentially one Mac OS X, but Canonical keeps spinning out slightly different Ubuntu releases?
I must concede, I’m not using Kubuntu, Edubuntu, etc. So perhaps I should give them a try before saying Canonical should trim back and focus its efforts…
-jp
Editorial Director
WorksWithU
Part of OS/2’s problem was that IBM was not 100% committed to it. There were factions within IBM that preferred Windows, or at least disliked OS/2. And that contributed greatly to it losing in the market. I don’t think the same dynamic exists at Canonical.
But you are right about its focus on Windows compatibility at the expense of native application support. Vendors felt no need to release native applications, because OS/2 ran their Windows products, often better than Windows itself.
But it was Windows 95’s APIs, which were vastly different from Windows 3.1, that were the death blow for OS/2. Windows 95 obliterated OS/2’s compatibility with the current version of Windows. As ISVs moved their applications to Windows 95, the catalog of current software available for OS/2 dropped like a rock, and its viability as an alternative to Windows dropped with it.
In that case what’s to stop Canonical from making a full Android distro if it grows as a platform.Remember this is free/open source software we are talking about.
I disagree. Comparing OS/2 to Ubuntu is absolutely apples to oranges. Canonical isn’t IBM and the Opensource World isn’t the Compercial World. Read more about OS/2 on wikipedia http://en.wikipedia.org/wiki/OS/2 you will find that there were many reasons why OS/2 lost.
tough situation that Canonical is in, a good situation, but tough. They have been arguably the first linux distro to really truly get the attention of the masses. Shuttleworth and co are doing amazing things are looking to build a business.
So, where do you focus. The fun part of startup mode is that you can adapt and move fairly nimble in the early days in order to find your groove. Need to find it soon. I love ubuntu and looking fwd to the future, whatever it brings.
Joe: interested read. But I think the main difference here is that both Moblin and Android are open platforms, so adding a compatibility layer (or working on interoperability with them) is a win in general for the open systems, and a loss for Microsoft and Apple (which are both trying to push us into vendor lock in, thus becoming my foes ;-)).
Let’s take a look at the Android architecture:
http://developer.android.com/guide/basics/what-is-android.html
It is basicaly a Java set of API’s on top of the Linux kernel. This is,in a sense, like supporting another desktop (besides KDE, XFCE, Gnome). And now that Java went GPL, why not 🙂
Same goes to moblin.
So, in short, while I appreciate the historical insight you are giving, I think you gave the right arguments for the wrong target, LOL. I think supporting Windows throw WINE is an issue, in part for the reasons you gave 🙂
My 2 cts. Thanks for opening up the discussion!
Thanks for all the reader feedback so far.
I realize Moblin and Android are “open.” But Canonical still needs to keep the developer spotlight primarily onto Ubuntu.
And just to clarify, I realize OS/2 lost because of numerous errors at IBM — and numerous smart moves at Microsoft.
But strong native applications have always been one of the key reasons certain operating systems win or lose.
I remember, when IBM were marketing OS/2 on the PS/2 I mentioned at a press conference that it looked like “half an operating system running on half a computer” I didn’t get any more invitations to IBM press conferences for about six months afterwards.
Hi Joe. I disagree, I think we all need to push for open standards and get people to write for the standards, and that’s a win for almost all, and a loss for MS and Apple. Canonical, I think, needs to try to keep a very large mind share in the Linux ecosystem, and capitalize on that (in the form of contract to customize Ubuntu for phones, netbooks, desktops, servers, etc). I think this is the natural way to go, but I could be wrong 🙂
@Bane:
Creating an Android compatible linux distribution is one thing…technically it can be done..that isn’t the issue.
It probably even going to be popular with Android developers as a workstation environment…a no cost workstation environment.
But building a business off of it is quite another matter entirely. Can Canonical build a business off of Android compatibility? For the case of Android support, building a solid business will mean convincing OEMs to pre-install a Canonical built Android clone instead of choosing the official Android by Google. How exactly does Canonical make an Android-compatible alternative more attractive than Android itself to OEMs? Its not like Android is a large cost burden to OEMs.
The business economics for Canonical are not driven by whatever secondary ooperating system the minority population of technically proficient users choose to install for themselves. Having an Android compatible Ubuntu will certainly be embraced by the technical userbase..as it will help them more quickly develop applications for Android. But none of that activity is actually a revenue generator for Canonical. What is the business model here?
The economics are driven by what OEMs pay money for in integrating a pre-installed operating system. And it looks to me like Canonical is setting themselves up for a major uphill effort to try to displace Android with something Android-like.
Yes, Android is open source, but Google still has a competitive advantage by defining what Android is with every Android release. Canonical will continually have to play catch up and make an extra effort to certify their Android comptibility for each revision of the Android platform spec.
I don’t exactly see Google going out of their way to help Canonical build a business by cloning Android. How do you build a business chasing specifications that only your competitor gets to define? Android will be a moving target.
The first generation of ARM based netbooks/smartbooks were demod at Computex last week…they were ALL running Android. That should be taken as a significant indication as to where things are headed and why Android support is important to Canonical. Canonical announced an ARM porting initiative last November…has any ARM OEM signed up yet to ship UNR for ARM?
-jef
Personally I think the biggest mistake Canonical has made is to try and take on Red Hat and Novell. So far as I know the vast majority of Ubuntu installations are desktop installs. Theirs a really opportunity for Canonical their that they just don’t really seem interested in taking seriously.
Jef: I know sometimes we don’t see eye to eye. But in this case, I think we’re on the same page: I think Canonical’s MID and Ubuntu Netbook Remix editions face uphill battles. Though our readers do remain interested in those types of Ubuntu devices, I think the device makers themselves are increasingly distracted by Android, Moblin and Windows XP/Windows 7.
Aikiwolfie: I still believe strongly in Canonical’s server push. And I think we’ll see some new hardware partnerships before the close of 2009. Just an educated guess…
Joe:
I dont see Moblin as a distraction. Moblin is basically a reference implementation for a new interface concept. I think of Moblin as an extension of Gnome designed specifically attach the problems of the netbook form factor interface. Where Gnome focus on the traditional desktop. They are not inherently incompatible and they share a significant amount of framework. Especially when you consider that clutter is going to be just as important for Gnome 3.0 as it is for Moblinv2. I expect to see Moblin and Gnome Mobile initiative converging rapidly in terms of roadmaps and shared effort on shared technology if not the full UI.
Canonical will be able to consume and adapt Moblin without a huge problem..just as they do GNOME. Neither Moblin nor GNOME are in “competition” with Canonical at the moment. Gnome project releases its own live cd reference images too..but its not a distribution competitor generally.
If a Moblin remix replaces the UNR thats currently available, that’s not a bad thing for Canonical necessary. But they’ll have to be willing to let go of their homebrew developments, which can be hard to do. Moblin beta images use PackageKit by default for its package manager. Canonical continues to invest in its own homebrew packaging UI solution for Ubuntu, instead of tasking paid time to polish off PackageKit as a common solution across distributions. Even the more community controlled Kubuntu has decided to bite the bullet and use PackageKit by default even though the .deb backend issues which Canonical cites as critical for Ubuntu adoption of PackageKit are unresolved.
That is the sort of problem Canonical is going to face when integrating Moblin. How far are they willing to diverge from the reference versus taking the time to invest manpower into the upstream development of the technologies that form the reference? Canonical was in on the ground floor on Moblin and it sort of got away from them. They may have lost their chance to heavily influence the technology direction Moblinv2 took taking…but they can still come back now and start to impact Moblin if they engage. The CouchDB work that underpins UbuntuOne for example is something they could bring to the table for both Moblin and Gnome as a deeply integrated networking data syncing facility. And they are really taking a lead in terms of desktop usability testing inside Gnome, they could come back into the Moblin development arena as a leader there too. Moblin is more opportunity than competition.
But Canonical has to engage in the Moblin roadmapping and development to be able to make a credible argument to OEMs that they are going to be a good engineering and services partner for Moblin based devices compared to Novell for example. It’s not even clear that Novell has an advantage at the moment. Canonical has existing relationships with enough device OEMs to have a good idea as to what OEMs really want. They are probably ahead on that learning curve than Novell is…even though Novell can position itself as a technical expert on Moblin in a way that Canonical can’t yet.
Very different issue than the Android issue…where Google is absolutely a direct competitor in the space in a way that Moblin/Linux Foundation is not. Canonical has absolutely zero chance of influencing Android and that puts then at a very real disadvantage in the servicing and engineering business. Android’s engineering and servicing market is Google’s to lose..especially when it comes to the ARM OEMs where Canonical doesn’t have an established presence as a reliable engineering partner.
The real flash-over-points is watching to see if Android on intel starts being adopted more quickly than Moblin can be adapted to ARM/MIPS. It’s probably not in Canonical’s best interest to see Android adopted quickly as a native intel notebook OS.
The irony is that Xandros may bring Moblin to ARM faster than Canonical can and be the beachhead that Canonical can follow into the ARM space just as they were for linux on the first generation of intel netbooks. We hear very little about Xandros, but from all accounts they had a better technology showing at Computex than Canonical did.
-jef
I don’t think there is anything to worry about. I meet lots of ol’ timers out there that thinks OS2 was the next best thing since sliced bread. OS2 was killed by Microsoft, quite intentionally because that is what Microsoft is likely to do. Check the BeOS entry in Wikipedia to see what happened to BeInc.
With respect to Porters five forces Andy Grove proposed the sixth force, complements, which is where the modern computer industry lies today. It means software makers and hardware manufacturers complement each other to the point that when there is better software the hardware manufacturers get a windfall, and vice versa.
Microsoft protects its territory voraciously due to its appetite for market. It will never share the pie with any entity. We’ll this is short sighted because as we all know there are various applications in computing, not all requiring interactive computers. Linux routers are an example.
This last category is what demonstrates the value of Linux, and other freely available OSes. The theory of market complements goes beyond the above narrow definition. In application you can increase the market size in one market by increasing the consumption of a complementary product in the other market. Free OS and related software means greater sales of computers and paid-for software that runs on the free OS. You see this everywhere from linksys GL created by Cisco for Linux enthusiasts to netbook class to the rationalization in the cell phone hardware industry that making it easier to have one or a few OS platforms.
Intel understands the markets complements models, which is why it has over the years tried to get some things moving in software much to the chagrin of Microsoft (cf: the Microsoft Way), which resulted in a hostile reaction by Microsoft. Intel tried to move the low level sound software and hardware processing but Microsoft bulked.
We could have been a long way away from where we are in computers today were it not for Microsoft stunting the growth and direction of the industry through its practices.
Yeah, I know there are people that go, “but we wouldn’t be where we were today if Microsoft was not THE driver,” the I say exactly. It’s the reason that old timers say that OS2 was the most innovative of all the OSes to date. Its unsurprising the level of mediocrity that comes out of Microsoft as it grows. I believe beyond the rationalization of workforce, there must be a rationalization of sense of innovation. Microsoft for its entire life has bought technologies to kill them, tying two dead horses together and hoping they run faster.
Anyway, after all that all I wanted to say is that Ubuntu is no OS/2. Canonical owes nothing to the man. Linux in general will continue on to blaze new trails creating new markets for the likes of Microsoft. And Microsoft will be the angry bull in the cage of its own creation.
Makes not that much sense this article.
canonical wins with server ubuntu’s, not with desktop, OS2 Warp had a lot of license issues, Ubuntu doesn’t have license issues.
So a cost for the company is the server cost on licenses, because iwe biu the license of the server OS as well services attached to it… Canonical garantee a license free server and you can only pay services.
so i dont see any problems supporting others.
this article assumes that OS2/Warp was GPL(it asn’t), that it was free(it wasn’t), and that it was evoluting as a GPL OS(it wasn’t)
we are facing a really different OS world now
I find it quite interesting how a “show how it can be done” demo turns into a “speculatory”, “ubuntu is supporting android” conversation.
The reason oil prices moved so high artificially was by speculation and greed. Stop the sensationalism and ask the questions. You may not receive the answers you want to hear, but that’s the choice a private company can make.
@Joe: From a “speculatory” perspective, you might want to look/review the REST of UDS information for some really interesting insight into the progress Ubuntu is making. You may also want to look at the information around Computex to see how well UNR is being received.
Jef, I see where you’re getting at but I don’t think that Canonical are powerful enough to influence which platform will become dominant. OEM’s like Android because it can be closed (Apache license) and they are the ones running the game right now that’s what’s going to be on MIDs and netbooks. What Canonical can do is to provide both worlds as value for users who like freedom and businesses that want Android apps in Ubuntu environment(desktops, servers). Basically, I think they are catching the last straw because obviously OEM’s want to control what you run on your computer and are ditching the ‘classic’ Linux for Android. With the exception of Dell all netbooks ran obscure distros for the sole reason that Asus or Acer wanted to have their own way. Of course it failed miserably because they are clueless of what GNU/Linux and FOSS in general actually is and that is community.
This article is a bit of a stretch IMO. Canonical does need to be careful not to spread themselves too thin, but taking the goodness from moblin or adding Android “compatibility” is nothing like the OS/2 craziness of the 90s. Moblin has achieved some really nice performance increases that dovetail nicely into Ubuntu’s roadmap (ie boot time). Also the Moblin UI, as cool as it is, still doesn’t offer “that” much functionality. However, it is based on clutter, which could be nicely integrated into MID or UNR editions. This is the beauty of FLOSS, the ability to mix and match what makes sense while participating with upstream.
Android is a little more orthogonal, but I can see some justification for it. It positions Ubuntu in a place none of the other distros are in. It sets it apart. Perhaps there is a market for android apps in Ubuntu, who knows!
OS/2 had an identity crisis without the full support IBM. The unholy marriage of Microsoft and IBM created OS/2 and after the divorce, IBM just piddled around with it. I remember one of the biggest selling points IBM pushed with OS/2 was that it could run Windows apps.
I’m not sure there is much of a correlation between Ubuntu and OS/2. Ubuntu has really captured the mindshare for the Linux desktop that OS/2 never did. Let’s also remember their push into the server market is very recent, and their position on the desktop will do nothing but help them there. The problem is that the home desktop isn’t a profitable area. Home users don’t buy support.
What they need to do IMO is working on a tight integration between the server and desktop with some sort of AD alternative that’s easy to deploy and administer. They don’t need to beat Red Hat at it’s game, but carve out a niche for itself and the corporate desktop (yes it’s there but they don’t focus on it) is a market that’s largely ignored by the major Linux companies.
If they keep their products separate and have their desktop and the same server offerings as everyone else, people will try it because of the name, and depending on how the support is they may gain market share. I just don’t see them beating Red Hat in the traditional linux server market. They’re positioned well to enter new areas of the market and I’m hoping they have the vision to do so.
I don’t see any connection between Ubuntu/Canonical and OS/2/IBM, but I sure do miss OS/2 and I’m sad IBM treaded it so badly.
It is true that “there’s essentially one Mac OS X”, but only because you specified “Mac”. Apple produces at least five operating systems: Mac OS X, Mac OS X Server, iPhone OS (used in the iPhone and iPod Touch), iPod OS (used in the iPod Classic and Nano), and Apple TV OS. With the exception of iPod OS, all of these use an OS X base. So their approach is basically the same as that of Canonical: much the same base system, but markedly different interfaces for different form factors.
@mpt:
Except of course.. Apple is essentially a hardware company, not a software company. They design and sale physical devices and part of that design is the operating system. They don’t have to wheel and deal OEMs to pre-install their software against competing OS options. They build the device exactly how they want including the OS. Apple’s even talked recently about designing its own chipsets to have even more control over the final device design for its devices.
Trying to cast Apple as a software vendor, is like trying to cast Ford as selling light bulbs. Sure the vehicles have light bulbs in them just like Apple devices have operating systems. But Apple is building and selling much more than just the OS.
This is markedly different than Canonical’s approach.
Canonical isn’t a hardware company and that makes a very big difference in the final analysis as to how Canonical can position retail products. Canonical has very little control because its the OEMs that decide the overall the device concept. Ubuntu is just a decision that OEMs make as part of a larger device design.
Because Canonical doesn’t build its own hardware the role Canonical plays in overall device design is limited. Not only limited, but also more complicated than Apple. Where Apple has to target a very narrow set of device specifications for the iphone or the ipods… Canonical has to make a best effort to try to support different OEM device specifications all at once to stay competetive as an OEM pre-install choice. Fall too far behind in hardware support and OEMs won’t be interested in paying for engineering and support services when they can go with Android and have working hardware sooner. At what point is vast hardware support commitment a game of diminishing returns for a minority OS vendor? Apple avoids the trap entirely by controlling both the hardware and the software as part of a coherent device design.
-jef