[Sat Jan 24 15:21:48 CST 2004]

OSNews has published some benchmarks measuring whether 64-bit binaries truly are slower than 32-bit ones and the hell broke loose in Slashdot. As usual, some of the comments posted there are knowledgeable and intelligent, and some others appear to be based just in prejudices and dogma. The author of the article, Tony Bourke, chose to compile certain applications (OpenSSL, GNU gzip and MySQL) with 32 or 64-bit optimizations using the GCC compiler on a Sparc Ultra 5 and then measured the performance. The overall conclusions of the experiment appear to support the idea that 64-bit binaries truly are slower than 32-bit ones. Of course, once we think about it, we may notice a few issues. First of all, what is so attractive about 64-bit platforms is not so much raw performance of the binary measured this way, but the fact that they provide a much larger address space. Sure, this may not be necessary for your word processor or browser, but there are people out there who absolutely need this in order to do their work. Virtual reality, special effects, film editing, weather simulations, large databases, scientific research and military applications are but a few of these activities that need the extra address space, but I wouldn't be surprised if heavy gamers also need them pretty soon. Second, the experiment has a serious drawback: it is widely assumed that GCC is not very well optimized on Sparc and 64-bit architectures yet. Thus, comparing the binaries built by the widely tested 32-bit GCC compiler with the 64-bit ones is somehow misleading. The same would apply to architectures such as MIPS too. A much better comparison would have been GCC on IA-32 versus GCC on IA-64, and even in that case the latter is barely tested and debugged. So, we should always take results like these with a grain of salt. Nevertheless, it stresses something that applies to technology as well as to politics and life in general: first, always check your assumptions, because quite often they will be proven to be incorrect; and, second, most issues are never black or white but rather a much less attractive hue of grey. {link to this story}

[Sat Jan 24 09:26:38 CST 2004]

Much has been said and written about the fact that Linux is free (as in beer), and how that is causing headaches to Microsoft and other commercial software vendors. However, it seems to me that a far more important characteristic of opensource is too often forgotten or dismissed: the fact that it allows more freedom to the user or customer and strips the vendor from its usual unlimited power over the product. Sure, I understand most companies are simply looking for an application that does what they need, and this is even more so in the case of small and medium size businesses. I can even see how this may not be a key factor for large companies that simply need some software to assist them with their regular everyday activities. Where this freedom factor becomes important is in two or three very specific situations: governments (especially non-US governments) that need to use the software for military purposes, governmental institutions (again, especially abroad) that have a clear interest in avoiding the vendor trap and companies that need a platform to develop their products on, such as Tivo. The last case is especially intriguing, since quite a few people still assume that it is much easier and cheaper to purchase a platform from any of the well established commercial vendors out there in order to reduce development costs. What they don't realize is how this approach has a few flaws: first of all, the customer has to shoehorn the platform into his own product to make it work; second, there is no access to the source code in order to make the customization any easier; and third, the vendor is still in control of fixing bugs or adding features, no matter how critical they may be for the deployment of one's own product. In other words, if a small company decides to build its own product on a platform put together by a commercial vendor, they'd better be ready to show a business case if they need something fixed. All this applies even more to the public sector, especially when it comes to non-US Governments. It shouldn't surprise anyone that countries like China, Brazil or the European Union have been promoting the use of Linux and opensource in general by their agencies, since there are some very clear benefits they can reap: first of all, they avoid becoming overly dependent on a foreign product which leads to obvious political weaknesses; second, using closed source software developed in a foreign nation has some clear implications when it comes to military use; third, opensource allows for a much more levelled playing field when it comes to learning the tools of the trade and allowing young students to become involved in the development of enterprise ready software; and fourth, it also makes it possible for local companies to be created around customized opensource solutions. This is precisely one of the reasons why I am convinced that opensource is unstoppable by now, no matter what SCO, Microsoft and other large companies do about it. However, does this mean that commercial software vendors are about to disappear? Or, as argued by SCO, that this ongoing commoditization of software will damage the US economy? It would certainly be ironic to see the most pro-capitalistic, pro-free-market Government in the world intervening in the economy in order to save the day for some struggling companies and roll back what the market decided on its own. In any case, why should software be any different? Why should we sit passively while a myriad other products become commoditized and make an exception for software? I still have to hear some valid explanation about that. {link to this story}

[Thu Jan 22 10:46:08 CST 2004]

Eugenia Loli-Queru has published an interesting article on Mono that is causing some stir in the Linux community. I still have to spend some time checking out the .Net, C# and Mono technologies but pretty much everything I have read so far praises it. As a matter of fact, I must say that most criticism I have read against Mono and .Net is based more on political/philosophical grounds than on technical ones. Not that this makes them less meaningful, of course, but at least it tells me that as a technology it seems to be quite sound. Sin embargo, Eugenia raises a few good points in her article.

C# is very easy to learn and GTK# gives a much better face to GTK+ (plain C GTK+ development seems to me pretty sad comparatively) as it makes it much easier & faster to develop and results to more secure applications. (...)

Mono has still to fight off a few Linux die-hards who don't want to see a Microsoft technology used on their alternative OSes but thankfully the situation becomes better everyday as .NET and C# put a stronger foot in the industry and start to get perceived as a "standard". Another "enemy" Mono will have to fight out (soon enough) is Sun. (...)

Someone could argue that Sun would be pushing for Java on Gnome too, however Sun has already done (yet another) strategic mistake and didn't create Java bindings for Gnome neither tried to integrate more to the Gnome desktop (despite their "Java Desktop System" distro, there is no major Java integration to the toolkit or the gnome libraries). Even their java theme wrapper for GTK+ themes is half-baked and almost no developer uses it. (...)

It's a pity, because Java is also a nice technology but it nearly seems as if Sun managed to waste its initial advantage just trying to keep sole control of it. By the time they reacted and opened up a little bit, it was too late, and now it is not clear anymore whether the future belongs to Java or perhaps to .Net. As for those who criticize Miguel de Icaza for pandering to Microsoft, the truth is that Microsoft's own efforts to open up the C# specifications and go through the standards process may be doing away with a lot of the original concerns. To be honest, I have doubts myself, especially taking into account what Microsoft has done in the past, and am still not sure that we should base GNOME on C#. However, if it does prove to be powerful enough in the end and the standards are fully open, I just cannot see anything wrong with it. {link to this story}

[Wed Jan 21 13:10:10 CST 2004]

The recent decision by Red Hat to discontinue its low-end Linux distribution is definitely causing a lot of trouble to sysadmins all over the place. Some are considering to pony up and purchase licenses to run Red Hat Enterprise Linux, while others are mulling over a switch to SuSE, Mandrake, Fedora or even Debian. The Fedora Legacy Project definitely looks like a fine choice to many of those who are finding themselves stuck with older releases of Red Hat. Of course, the question now is whether or not this project is fully capable of maintaining old releases and backporting security fixes but we have little choice other than wait and see what happens in this front. Since support for Red Hat 7.x, 8.0 and 9 has been discontinued recently, we should be able to find out soon enough. Still, it is interesting to realize that at least we have a choice thanks to the fact that this is an open source product. If it were a proprietary operating system and the vendor had decided to pull the plug we would all be damned right now. {link to this story}

[Wed Jan 13 22:25:02 PST 2004]

Red Hat announced today that it would contribute ownership and maintenance of eCos to the Free Software Foundation (FSF). Sure, the project had been basically on hold since all eCos developers Red Hat inherited after they purchased Cygnus were laid off back in 2002 or so. Still, it goes to demonstrate that Red Hat, contrary to what so many naysayers assert, is a very honest member of the opensource community. One may dislike this or that move, may feel bad about their decision to drop the Red Hat Linux distribution, and may even disagree with certain technical aspects of their product, but it's about time to acknowledge that this is no Microsoft like certain Slashdotters cry out there. {link to this story}

[Wed Jan 13 21:35:18 PST 2004]

I have been quite busy lately, so I didn't have a chance to write much here. First, we had the always busy and family-oriented Christmas holidays, then I had to put together a presentation on the SuSE init system for some SGI training sessions we had, and this week I'm in Mountain View on a company trip. Such hectic schedule didn't allow me much time to sit down, relax and put together a few lines. In any case, needless to say, life goes on as usual, and the technology field continues leaping forward. LinuxWorld publishes a story today on Novell becoming the largest Linux vendor in the market. Who would have said just a few months ago? Yet, after they purchased both Ximian and SuSE, Novell is perhaps the Linux company with more chances to build a truly integrated and successful Linux desktop that gives Microsoft a run for its money. It is, after all, what most geeks have been dreaming with since at least the mid-1990s. I have used both Ximian and SuSE products in the past, and must acknowledge that I got a very good impression from both. Generally speaking, if there is something I dislike about Red Hat is its sloppiness. Sure, after years running their distribution one gets used to all the quirks and peculiarities, but there is still no reason why the Debian developers could come up with something like apt while all the guys over at Red Hat could show was a lame up2date... and let's not even talk about the overall feeling of their product. There is just something there that doesn't fit, even though it's difficult to put one's finger on it. On the other hand, Ximian managed to develop the most user-friendly Linux desktop out there without any doubt, and SuSE put together a solid distribution that is very well designed and gives you the feeling of being cohesive and integrated, both characteristics that Red Hat definitely lacks.

So, it seems as if a real competitor to Microsoft in the desktop market is really starting to take shape, right? Our friends from Redmond should then very worried, right? Well, yes and no. On the one hand, they should be worried to see the Novell-Ximian-SuSE alternative forming but on the other hand it seems obvious that the old UNIX vendors are the ones who should be really afraid about the Linux wave. The Investor's Business Daily published an article today about how UNIX firms are crafting a plan to fight the Linux threat. And yet, even after reading it, one notices right away how there is something wrong with the whole concept: these firms are just reacting to the Linux advances. How could anyone win a war that way? Let's be realistic. The UNIX vendors are not so much crafting a plan to fight Linux as defending themselves as well as they can in order to gain some time that may allow them to figure out a way to incorporate Linux into their business plans. That's all that's happening. Every single time I read one of these articles written by business types, it seems pretty clear to me that they miss one important point. They pay too much attention to the fact that Linux is free and it runs on cheap systems. I suppose it's something we should expect. These are business people after all, and something that is free will definitely draw their attention. However, paying so much attention to something completely secondary, they seem to be missing the forest for the trees. Linux's true accomplishments are twofold: on the one hand, it is free as in freedom, releasing the customers (especially big firms) from the submissive relationship to a single vendor they have learnt to hate in the recent past; and, most importantly, Linux provides a common set of APIs that, while UNIX based in nature, represents the standard that all these UNIX companies destroyed back in the late 1980s and early 1990s. It's no wonder then that Linux poses a major threat precisely to those UNIX vendors of yore. The colorful rivalry between Linux geeks and Microsofties is cute for the headlines, but the real loser here is UNIX as we have come to understand it in the past two decades or so. In this sense, companies like Sun have more to lose than the likes of IBM or even SGI, mainly because Sun has been too slow to react. Sure, that will still not save AIX or IRIX, and I still believe that Sun will be the last of the big Unices to fall. Yet, I am convinced that while IBM (and perhaps even SGI) will manage to flourish by riding the Linux coattails, Sun will have to face some tough times. {link to this story}

[Fri Jan 2 11:38:00 CST 2004]

It seems clear by now that opensource has gained enough respectability to be considered as a serious option by most companies. Who would have thought just a year ago that Sun would be releasing the Cobalt's custom user interface and BIOS under an opensource license?. Sure, they have also decided to discontinue the product altogether, so it is not as if they are opening the code to Solaris or anything like that. Still, not so long ago Scott McNealy and other Sun big shots could hardly hide their despect for anything opensource. Today, not only are they making a Java Desktop System that is nothing but a Linux box with some customization on top, but they are even releasing some of their code under the BSD license. It is a clear step forward, I think. It just makes sense. If a company decides to discontinue a product altogether, isn't it wise to opensource the code so at least customers can continue fixing bugs and scratch their own itch? Actually, with some luck an user-centered community may develop around the product and ease the pain a little bit to those who found themselves all of a sudden with an end-of-life product from a major vendor. {link to this story}