Technology and Wisdom = Better Humanity
We should be wiser in using technology for human happiness.
Monday 4 June 2007, 3:35 AM
Strengths of Open Source opposing Microsoft patent threats
Recently, two actions by Microsoft seriously called the attention of the open source community. One is the strange alliance with Novell and another is the unproven statement that GNU/Linux infringes on 235 Microsoft patents. Although the community has the feeling that these actions are outrageous and illegal, it does not know how best to deal with them, since in the U.S. one has to spend much money in order to respond to unfounded statements by anyone, especially when the opponent is very wealthy.
I have the impression that open source to challenge it in a justice court is exactly what Microsoft desires, so we should not get into this dangerous and expensive game.
Therefore, I propose another game, in which the open source is strong where Microsoft is weak.
Let us organize an event, where we place two large tables in a convention room. At the head of one table, one places a server hardware running Windows Server 2007 prepared as a thin client server with well-known applications: MS Office, Adobe suite, Interne Explorer 7, Outlook, for example. At the head of the other table, one places the same server hardware running a GNU/Linux distribution (RedHat or Ubuntu Server) prepared as thin client server with open-source applications such as OpenOffice, the Gimp, Firefox, Evolution etc. Along the tables, as many client PCs as possible are connected as thin clients, so that all sorts of people would run applications on the servers simultaneously, trying to overload the corresponding server with their input. On the wall there will be a projection showing how many applications are open on the server, and how they are loaded, like htop on GNU/Linux, so the surrounding people and journalists can see, and document, the load changes, the memory load, etc. in real time. It would be more fun if external users be allowed to log in to those servers and participate in the effort to use intensively the applications.
My contention is that the Windows server can take less load from thin clients because the applications cannot handle many users, while the open source applications are able to handle many simultaneous users without crashing. For this reason, if many users open MS Word on their terminals, the server has to launch many copies of the application thus filling memory and going for the swap space sooner, eventually going to thrashing mode.
I remember that this kind of test was done before, but never in public, side by side. This test will be entertaining if the important people of both sides show up, address the public in keynotes, and exchange ideas about the ongoing contest.
The objective is to show for all to see how lean and mean GNU/Linux is as compared to Windows, and demonstrate that they are so different that GNU/Linux cannot be infringing so many patents and behave so differently. It is impossible to describe such differences with words, so we need a great, visible show. I guess people would like to buy stuff there, and there would be less waste of time and money than in a courtroom.
PreviousComments on this post
Sounds like it would make a great spectacle. I wonder whether we could organise something like this...
Thanks for the favorable comment, Matt.
To add variety to our show, we should have a workstation loaded with various virtual machines running all Windows OSs including Windows 3x and 9x, and various GNU/Linux distros, so that we can compare how fast they run under similar conditions on the same real hardware. This would be a show I would like to see.
Thats sounds like a great idea, I would be very interested in attending such a show as I am sure a lot of other people would.
It would be interesting, to see how the two Operating Systems stack up side by side, and it would also be interesting to see how each of the previous distributions compared to each other.
About location: Could be somewhere in Europe, because in the U.S. there are too many people who would try to depict the Microsoft products under a favorable light and bias the results. Would someone suggest a good venue for the event? The requirements are good space, not necessarily huge, with plenty of internet connection facilities, and not too expensive, which would indicate a secondary city, or a city trying to attract an important event. It might have the business significance of an international sports event. However, I have nothing against any other continent; it is up to how many interested parties there will be.
About documentation: private camcorder owners should capture as much as each can, in order to cover the event completely, even more than a car race.
People at home would watch through the servers in question, participate in loading them with tasks, and capture screens for documentation, with timing etc.
The resulting movies with comments would be made available worldwide (utube, bbc, hollywood, whatever), and eventually would be put together into a 75-minute documentary. This movie would be a hit, something to be talked about for months. Its publicity effect in favor of GNU/Linux would be much more significant than the non-hostility agreements Microsoft is achieving (Novell, Xandros).
About timing: Autumn is good, or Winter, perhaps, but still in 2007.
About organization: I assume that many people will have interest in appearing in the movie, testing the setup in loco, or sell something. Once we decide on the venue, and estimate the costs, we could establish a registration fee for participants, visitors, journalists, booth-holders, food-and-drink boxes, etc.
We need a local organizer team to deal with such details. I wish the event to be self-financed, if possible; it will be a non-profit activity--our reward will be the results. If necessary, one could start an Open-Source Testing Foundation or something like that to be able to handle the money, the movie-editing, publicity, etc. I suggest that if there are profits, they will benefit open-source initiatives in some way. The team will decide who will provide the hardware, the Windows licenses etc., too. Always remember that this is a grassroots movement providing a showy response to proprietary efforts to render open-source meaningless. All who have benefited from GNU/Linux and FOSS have an opportunity to contribute for the success of this event.
Call to action:
I request readers to think about the deep implications of this initiative. Some town could become famous for hosting this event. If any reader wishes to attract this benefit to his/her country, please start your campaign immediately. I hope ZDNet.co.uk will spread the word worldwide.
While I think your analysis of the Microsoft patent hoopla is shallow but that is of no concern because the matter is nothing more serious than annual summer FUD. (Note that the Microsoft's announcement was a great boost to the adoption of GPLv3, and since Microsoft is a distributor of (Novel) GNU/Linux there is reason to think that the GPL virus has infected Microsoft and made the patent claims null and void. In any case Microsoft can't sue its customers without really making them angry, and it largest customers are also using GNU/Linux so Microsoft can only make FUD noise. (there is a significant threat from small holders of patents and they threaten Microsoft's big pockets more than Free Software's non-existent pockets. In actuality Microsoft and Free Software are going to end up on the same side of the US Patent issue.)
I do however love your idea of matching Windows Server 2007 against a GNU/Linux server. But lets do it a bit more scientifically, as well as ad-hoc. Lets match the systems thin client for thin client with thin clients matched on the respective networks and assigned the same specific task to do that will run automatically to completion once started. Advocates of the respective OSes will be allowed to choose what they consider the best software for the given task. A variety of tasks should be involved. The test should be run several times: Test one, start all the tasks at once. (what kind of load can the network endure and how will it handle it?) Test 2 start them one after another with a set time interval before the next task begins, the intervals between start times could be different so long as they were the same for the same task. Interval times to be determined by having each server do the task without any load and averaging the time the two servers take to complete the task. (If one network or the other has an ap that simply is not the equal of the other network's ap for the same task this will be revealed, though it should be agreed ahead of time that if an ap is incapable of a task the other network's ap is capable of, the task will not be included in the competition.) Test 3 Every thin client running a repetitive word processor or spread sheet "load", every thin client having another window open with a unique task to run, each unique task to be completed before the next thin client begins the next unique task. (steady load with unique load added. Can count the completions of the repetitive load as well.) Test 4 Windows and GNU/Linux geeks with expertise in the ap they will be using to accomplish a specific task will be pitted against each other. Experts will be given the task before the test so that they can figure out how to complete the task as efficiently as possible before the competition. Everyone will begin at once. This will match experts of presumably equal ability doing real world tasks on the respective thin client networks. This probably most accurately matches real world environments and at the same time equalizes the tasks and task demands on the system. The variable will be the software. May the best network win.
I'd also like to see a test of Vista against several GNU/Linux distros.
freedom:
I like your reasoning. Would you like to prepare a draft of the rules of the experiment, based on what you wrote? The draft would then be downloaded and checked by anyone interested in improving the plan, and returned to you for finalization.
That would result in a schedule the activities to be given to the executive team.
I think that one day would be enough, starting at nine and ending at five, unless we want to test the effect on the servers of prolonged activity, until memory leaks cause measurable slowdown, for instance. In this case, maybe two days would be necessary. We need more people to think about this and other details such as which applications we should test.
Regarding the test of Windows Vista vs. GNU/Linux, we need to make sure that they run exactly on the same hardware. We need to choose such hardware. I think it should start with installation of each OS, with people documenting, measuring times, then activations and application installations, then playing media and desktop effects (Aero vs. Beryl, Windows Media vs. MPlayer or Totem the same movie, etc.) according to a careful list of activities.
There might be unpredictable things, like a server crashing; what would be the procedure in such a case, regarding time measurements? Stop the clock?
As you see, we are going to face interesting problems of logistics. We will gain significant experience in this kind of, should we say, public benchmarks?
Again, freedom, thank you for your input. I pointed at this thread to the discussion group of the FSF and I expect input from them, too.
In the meantime, there should be people thinking about becoming the executive team, which should be in the same region as the chosen venue. Possibly the UK?
Yes, I will give thought to the rules and come up with something more fleshed out. I am taking the task on with the belief that the competition will actually happen. I am soliciting help from members of the local Free Software Association I belong to. I may or may not post anything concerning this for a week or more. I assure you that not posting does not mean lack of interest or thought. I will check to see if suggestions are made here.
Regards the desktop Vista vs GNU/Linux distros competition, I suggest the same machines used in the server competition be used. That would only require the additional expense of hard drives to swap out, and would simplify coming to agreement about what machines to use. The machines should probably be commercially available machines with "Vista ready" stickers on them, and in any case need to be machines both the Microsoft team and the GNU/Linux team agree are fair machines for the competition.
One issue will be non-free software, especially drivers. Personally, I think the competition should include totally Free distros as well as distros stuffed with non-free software. The Microsoft team will cry foul if we don't allow good to excellent video so when totally Free distros compete they will have to be allowed to install video cards that have totally Free drivers. If nothing else this will point out one of the problems of non-free software. In these cases two tests should be run, one with Vista having to use the same card the total Free distro is limited to.
Thank you, freedom, for the support.
Windows Vista vs. GNU/Linux: although it might be expensive, I suggest that two equal PCs with Vista preinstalled be acquired from Dell, and one of them would double-boot GNU/Linux for the occasion. After the event, these machines could be given to visitors through a random drawing. That would be a special attraction to increase participation. An alternative would be two of those computers from Lenovo or Dell without and OS preinstalled. Then a fresh installation of the OS would be the first stage of the contest. Which brings another question: Vista Ultimate?
Well, let us keep thinking on the details.
Sponsorship: I think we should ask for sponsorship and presence from companies such as AMD, Intel, nVidia, ZDNet, Dell, Lenovo since they will be involved in the competition. That will allow us to make it a more splashy event.
I have good news.
I posted our link on LinuxToday
here
and some people commented positively. One of them, Jim, said that his LUG (Linux User Group) would accommodate the event soon. So, I request freedom to contact Jim through that link to share your plans, or simply post it there.
As someone suggested, I tried to send a summary of our plan to Venezuela, through a site VenezuelaTuya.com, but probably I failed, because when I hit submit there is a message that says "Please wait..." forever. Well, I tried.
there are two good reasons why os/x should also be included in any public bake-off ...
1) since osx is based on darwin (mach+bsd+etc), it also might be effected by microsoft's (wild) claims against opensource in general (not just linux in particular?).
while there are obvious differences in license regimes bewteen darwin (and bsd itself) and linux, mircosoft's claims are so broad that it is not possible (?) a priori to know for certain that bsd or darwin are excluded from its scope ...
... of course it is more than ironic that apple - preNeXT - was already shipping its own unix decades ago (via AU/X), as was microsoft itself (via xenix); so mircrosoft's claim prior of art viz a viz windows (as distinct from derivative works) is of dubious merit prima facie.
2) any benchmark from (let alone a documentary about) the bake-off would almost inevitably confer lots of bragging rights ...
... to exclude one of the three big "little iron" platforms would not be a fair or accurate representation of the real-world environment.
in other words, there is no corollary that requires linux qua linux on "big iron" just because "big iron" doesnt run windows or osx: benchmarking linux running on everything from ipod to mainframe is obviosuly irrelevent to a bake-off.
If x86/64 hardware is going to be used in a bake-off, then the serious/mainstream OS's for x86 should be used - which means os/x is in not out!
NOTE: because some (opensource) apps are not pre-optimized for the mac, it is only fair that the configurations used for the tests be set-up properly for each platform by domain experts ...
there is the notorious example of an os/x benchmark for a scientific app that used the worst possible memory allocation settings and thus purported to run nearly an order of magnitude smaller, which was deemed to be the performance penalty of a microkernel design - when that is clearly not the case!
::zahadum
