Surprises me that no-one has figured out the real reason. Its commercial.
India's already trying to take some of the Euro satellite launching business away and this only helps further technology in that direction. Groups of developing nations might want to share satellites which India or China would launch for them. O pay to have their own sent up.
Would be good too for developing spy tech, to spy across noth countries various borders. And there will be missile tech spinoffs too.
Agreed entirely with the author
about the malaise in Linux desktopland. Infact, I wrote a not so well
proposed article about it almost an year back: Whats Wrong
in Linux DesktopLand.
Red Hat has since attempted to do a unifying job, but the question I
ask is this: how many of the applications that Red Hat ships will they
support? As an end user, why ought I have to deal with figuring out how
to install mp3 playing or DVD playing capabilities. Why should I not
have good fonts? This is the value add commercial outfits are supposed
to provide.
But they cant because they are too busy making too many CD's. Why not,
as the author has suggested, pare down the number of applications, and
pay a royalty to the author/maintainer of the application for each copy
sold? Why must a Linux company follow the same path as a standard
company in not renumerating the author? (To be fair, RedHat employs
many application or subsystem authors, but why not pay the others a
royalty instead. For example, why not pay Ximian for eg a royalty to
maintain Evolution to be consistent with RedHat design guidelines?).
The authors suggestion of one toolkit is important too. I applaud
Lindows and Lycoris for dumping gnome and making KDE based applications
(though I dont applaud everything as root idea..why not use
capabilities and gradually eliminate root from most applications). I
dont agree with KDE as a toolkit choice as the high licensing cost of
Qt screws small developers wishing to develop commercial apps or
shareware. The Mac is a thriving desktop platform precisely because of
these people, and we need to attract such development if we want to
keep the long term viability of Linux..dont forget that windows started
out as a poor desktop implementation, and but for linux+bsd's would
have largely wiped unix out of small and mid-range installations.
LGPL toolkits are good choices...
Here's one possible plan. Create a new distribution, I like to call it
birdbrain because thats all the brain one should need to use it.
Elitists not welcome. The basic subsystems are kernel+device, init,
basic unix utils, binutils, libsystem(libc, curses, etc), directory
services/auth. Thats 6 subsystems..create 6 teams, and assign
royalties. Get basic X. Pay royalties. Get basic languages: perl,
pythonChoose the basic desktop, say gnome. Get Ximian to package it,
and get Ximian redcarpet to distribute it. Pay royalties. Choose no
more than 15 gnome apps as part of the basic package..choose teams for
each, hopefully including original developers, who are willing to fork,
customize to needs of distribution. Needs are for (IMO): browser,
instant messager, email/news, news aggregator, editor, wysiwyg html
editor, rdesktop, file manager, package manager/installer/redcarpet,
media player, pdf reader, terminal. Thats all. Make sure media player
can play both DVD's and mp3's. If this requires factoring licensing
costs into the distrib, so be it. USABILITY comes FIRST. Then choose
personal server apps for fileserving, personal web serving, ssh
serving.
Thats it in the basic system. If this sounds like taking a page out of
Apple, well, yes it does, except that the whole commercial idea here is
to get money directly to the developers who maintain the app for the
distribution. Think of it as debian on a much smaller, and thus way
more coherent scale.
Now make add on packages with separate royalty and responsibility
spheres, for development(compilers), science(plotting, etc), office.
Anf of these packages, and also the previous 15 odd apps, ought to
replacable by others provided they provide the same task capabilities.
Nautilus can then be made more task oriented too, where tasks are done
independent of the apps providing them.
Create an experimental distrib in which new things are played with
before being dropped into the stable distrib. Examples would include a
unified way to deal with data in text form like Apple's plists or
RedHat's al
I'm not sure I understand, but modulo shared libraries, dosent having two loaded GRE's increase the memory bloat at the expense of disk bloat. Last I knew disk was much cheaper than memory...
Phoenix extensions would seem to be the only way to go then..but why are phoenix extensions better than standard mozilla ones?
Everybody seems to be missing the larger point here. This capability is the stepping stone to being able to put up spy satellites. In a few years India will be able to keep tabs on Pakistans cross border infiltration and nuclear installations, and hopefully, in the case of fundamentalists taking over( which I dont think will happen, Pakistan being to a large extent a fairly sensible country), be able to make a pre-emptive strike.
If you want to learn more about the origins of this programme read Abdul Kalam's Wings of Fire. Its a very inspiring book. That Kalam is now President of India(which is a titular position without much power, unlike the Prime Minister), is itself a testament to where self-reliance and competition in science and technology can take one.
Hopefully the programme can now be commercially self sufficient, and the pace of space exploration and missile defence research becomes faster. As you have probably realized in the last year, South Asia ia a tough neighborhood: a dictatorship to the west and east(Pak and Burma), the worlds largest communist state to the North, and ofcourse, central asia and the unstable 'stan's near by..
Lastly, such development can only serve as a long term counterpoise to scary go-it-alonists and US supremacists like some members of this administration...
I work at a university, and I can see clearly who is switching.
Those who say they wont switch here are probably system administrators. Since I do sysadmin as part of my job, I can say that that part of me is a control freak, and loves the power of linux. That is also the reason why Linux has it hard on the desktop: only macosx, lycoris, lindows are even thinking of deprecating root in their OS'es.
The part of me which programs is split. Doing scientific programming today is easier on linux because of the number of high quality numerics/graphics libs available for X11. This will change. However, have you seen the simplicity of macosx? Every app is a directory. No gtk compatability problems(for those who remember). Copy the app anywhere. click, go. For command line people, change defaults using the default command, since all apps use plists. Open any file by saying open bla.pdf. It will use the default app. use open -with if you want a specific app.
The linking model is simple. The loading model is simple. applescript scripts most apps and is way easier to use than COM or bonobo. Still linux is a familiar model to lots of people. So I know people now, grad students and post-docs and engineers, whose desaktop is a macosx box and who program on linux..the professors dont program much so macosx works well for them. This student/scientist/engineer/programmer is the only remaining market.
But at the end of the day its the apps. Excel is available. And itunes and iphoto just rock.
There was a time when i liked struggling with linux to get all this working. At some point, one just wants to code. One dosent want to deal with dependencies, etc. You will say apt-get and I'll say hallelujah, its a great thing, but why cant i just install the freaking app where I want it too, and delete it by trashing it. rpm --erase??? Who would think of that?
The sad part is, most of what macosx has done could and still can be done on linux. Make a restricted distribution. Share earnings with app developers. Choose 10-15 best-of-breed apps, thats all. Thing of the next evolutionary step in these apps, rather than remaining behind the curve. root should only be a single user mode thing. Like gentoo, make init scripts dependent on whats running and whats not. Simplify the runlevels to single-user, and multi-user. Reduce hardware complexity by certifying systems based on linux friendly manufacturers. run daemons not as root. Get rid of the start, or hat, or whatever menu. Get rid of the XP like icons(see redhat8 beta). Give gtk a default look which dosent look like grey shit. Use a tasteful muted color scheme. Make sure pcmcia and usb and firewire just work on plug in. Use hotplug and devfs like mandrake do. Get rid of one million etc config files and use gconf and alchemist like redhat do. Simplify the gnome2.0 desktop. Check out the innovations in oe-one's desktop. Use autofs pervasively. Implement per process namespaces. Implement a simple event layer on top of bonobo, pipes, mimetypes, clipboard, etc to make scripting the desktop trivial. See plan9's plumbing. Unify zsh(bash) and nautilus to use same mime system. Allow apps to be manipulated as directories. When such directories are opened in either, allow hooks to be called which can start or install apps into a dependency database. Create a pasteboard server like in macosx. Implement gnustep over gtk2.0...
You get my point. There is so much thats already there but just missing a bit. It needs people with that extra bit of innovation, and that extra bit of compansation a app-royalty scheme would generate to push it across the edge. It needs that part of me that is a system administrator to let go. But it may be too late.
Exactly why isnt Michael quitting slashdot then, since slashdot is the spawn of the evil proprietary software producing VA.
Sheesh, talk about integrity.
And while we are at it, if a proprietary piece of software with an OPEN FILE FORMAT was used to get me my refund or license or whatever faster, so be it. Let the people who actually do the work choose the best tool for the job.
Combining forces is a step in the right direction. But how can 4 companies survive competing for the same small services market for servers, where you are selling to system admins who know what to do anyway, and will buy only hardware related support? Yeah, some will sell as insurance, but I'd rather buy my insurance from IBM who will be around tomorrow.
The only market that would have bought in droves, and did between 1999-2002 is the academic and technical workstation market, where the need was plenty and the expertise thin. Where was the user centric product then? Today's linux companies are making the same mistakes that the unix guys did, leaving the desktop to be picked up by microsoft, concentrating on the server. And today, OSX is replacing linux as the desktop of choice for power users.
And whats with 1000 packages with 10 email clients, all substandard? Why package 2 desktops? Make a courageous decision and pick one! Why duplicate work, decreasing productivity? Do RedHat and Ximian need to both package gnome or evolution? Why not contract it out? Why not pay the gphoto developers a royalty, something you can do if you had just 15 desktop apps? Linux is presently sustaining programmers through VC, not through profits. This isnt a get-rich-quick scheme. Support the developers. Provide user testing for them. Give them a chance to live atleast part time of their software and consulting.
But most of all, dont leave the desktop. For they who dont have the desktop today wont have the server tomorrow.
I wrote some more about this stuff, see http://3point0.nareau.com and the 2 links on that page. Also see http://reno/~rahul/venn.jpg for an example of how an ecology of companies around a linux distribution and an application server(spacestation on desktop and cloudserver on server) could work.
And email me at tig@nareau.com if you want to do something about creating a distribution with one desktop(gnome as gtk is lgpl, i believe in letting developers choose their own license, so no religious nut jobs pls), few well done apps, attention to quality, user interface and simplicity rather than emphasis on service contracts, and a biz-model by which work is distributed to the actual package developers on a per-product sold royalty basis, and selling the software is supposed to bring in money.
I am concerning myself basically with the desktop. Personally, I dont think linux is doing badly on the server now(but see down), though I question the need for 5 not-so-heavily-differentiated distributions.
I wrote an essay on this late last year; its at
http://3pointo.nareau.com/stories/storyReader$5
One question worth asking is after the 3 years of VC 98-2000, why do we still not have a decent set of true type fonts, free.
The point is, the desktop was abandoned by existing companies for the server. Seems to have been a sensible notion by the companies?
NYET! History has shown that dominance on the desktop today leads to a dominance on the server tomorrow. Unix already dropped the ball in the early 90's. We are doing it again.
Lets get this clear first, Nautilus is a great and beautiful product. In my book both are important.
In March 2001, Eazel laid off 40 employees, paring itself down to 35. Even at its height, there were 13 out of 15 developers on Nautilus paid for by Eazel. So there were 60 other people in this company. Ok so estimate 5 execs, 5 graphic designers, 5 marketing and sales(they werent selling anything yet), 3 sysadmins, 7 customer support, 5 administrative, and we get to 30 more people. Obviously at no time did they seem to need to be more than 45 people (though 3,2,2,2,2,2 =13 more = 28 total which is what they got down to seems more realistic).
They got 11 million in March 2000, 1 year at 100,000 per employee for 75 employees ate 7.5 million straight, leaving 3.5 million+founders investments. At 30, they would probably still have had another year to try and validate their business model, or be acquired.
So I get the feeling that their burn rate was too high, too dot-commy, and that lead to their going under. Of course its easier to say this in hindsight, but they probably expected to clinch a second round easy. They should have been warned by last april's crash.
There are now many companies with embedded versions of linux. Most of their web sites have no provisions to give me the source.(Given that really what these companies sell is a config.mk, thats ok I guess). But a lot of these companies make changes I would think to make linux small, include special functionalities. And without providing a listing of whats free and whats proprietary module add-on how am I to know that they are not cheating the GPL(too many me-too companies out there for the nasdaq buck these days..).
Attitudes like Calderas have ensured fractionation. Some of the lack of cohesion in embedded development is tailoring for different purposes I am sure. But see how nicely linux developed all these years, with the older companies(even caldera) playing by the rules. Qualitatively, I get the feeling things have changed.
I'm telling my friends and family that caldera is a bad buy(software and stock). They are a company without a focussed market thrust anyway. First it was openlinux desktop, now it is some e-crap, tomorrow it will be something else. RedHat, Debian, and Mandrake(distribs) and VA always give back their efforts to the community and are not leeches.
If you look at the page carefully, you will find that the money is being put up by the US Government through the National Labs. So the US Govt is now *directly* funding Open Source. The products would be used to further parallel programming, beowulf, and just plain normal programming in a significant way. Yippee!
Why Python? A lot of developers at Labs are Biologists, physicists, chemists, weather folk, etc who develop large beouwulf and other numerical codes. Given that most scientists pick up programming without any formal training, they are more interested in spending time on their coding and science than in Makefile and autoconf arcana(and dont tell me that a m4 based config system is nor arcane).
Python is easyly understandable. A java/C++ programmer would pick it up in 2 hours. A C programmer in a day or two. People without any formal training pick it up in a week or so. existing C and C++ and Fortran code can be easily wrapped using SWIG--this has been used to implement the outer calculation and visualization loop for massively parallel molecular dynamics programs. See http://www.swig.org,papers, for details.
This is the basic point: How do you extent the advantages of software enginnering to those who are not software developers, but simulators and scientists, and other engineers. And how do you stop being elitist and lower the barrier of entry into Open Source for the nonexperts and the common man. And how do you ensure extensibility. The answer is scripting, and good coding practises for the non-expert demand python.
This was exactly what I thought when I heard of the Sun Ray. I think they made up their own protocol, undocumented, though. Two ways to do this on Linux: 1)Create Protocol. Use Xvfb virtual framebuffer server and use protocol to communicate 2)Use xvnc X server, use linux svgalib vnc client, or write a new vnc client for the linux framebuffer. Someone had linux svgalib client on a floppy. Pop into a PC, boot PC, and you have a walking thin-client. Could even store a private key if needed. Replace floppy by card. Done. Tunnel vnc over ssh for encryption and compression and use cards to provide private key and encrypted password sent over network. (Second layer of protection)
Second layer could be, desktop linux PC's with CODA used to hoard releavant parts of linux servers. Could be laptops. VNC clients could still connect to these machines. Such a laptop would solve the problem of server downtime and use from anywhere in the world where there is no bandwidth. Companies may not want it(secrets). However, it could be used for workgroup synchronization and multiple options in case of failure.
Anyone want to do this? Email me.. (rahul@reno.cis.upenn.edu) Suns offering is really quite simple and we could have a product free in software and $400 in hardware INCLUDING monitor, utilizing new PC's, or people could use their old PC's as NC's. I call it Sting Ray!
The LDP docs, in the old days of Sunsite and few or no RPM's and debs, were crucial to anyone doing Linux. Times have changed, and the LDP (IMHO) needs to organize documentation btter I think.
Most People do work in the context of a distribution these days, and distributions do things differently. I've often found files at different places than in the docs. The docs maybe too detailed for me to spend my time on...I want to code, not administer my systems, nor compile everything from scratch. What I am saying is that the granularity of the docs is not fine enough--I'd be happy to know how to use Linuxconf to set up routing with a small sidebar as to what is going on behind the scenes. I dont want to read the whole Net-3 howto, unless that is my interest.
Finer Granularity implies better structured documentation, in the sense that it is not enough to say "section", but section on routing with references to the route giving the error message set..., so that I can zoom in on exactly what I want. The more general point here is that finer granularity needs to be intelligent granularity, so that searching is smarter, and natrual language searches are guided.
Now Docbook provides a lot of these tags. You can write API and command man-pages, there are tags like command, syopsis, errormsg, funcsynopsis, screen, screenshot, input(what the user inputs) etc. Some of these tags are common to all documentation, some ought to be removed, and be put into domain specific XML dtd's, like those involving class framework API's, etc
Docbook is too big to use. A whittled down set of tags, finer grained and more functional than linuxdoc, combined with domain specific XML dtd's which can then be combined with docbook XML using namespaces ought to be developed. What would have to be agreed upon, or volunteers found to tag, is a minimal docbook subset, and a method of combining DTD's. Conforming to these specifications, developers would be free to create their own DTD's, or convertable doc formats, and we would all reap the benefits of structured documentation.
Finally, from a users point of view, there is the documentation finding chaos. Umm lets see do man bla, is that fails, and even if that succeeds for a FSF software, do info bla, umm we're still not there, wait there is/usr/doc, and stuff is in html there, umm no its in PDF/PS, or oh, in tex and I need to process it....Now obviously this chaos will remain as everyone has their favorite formats. But searching is out of the question, and at the very least, we need a RDF source(like newhoo) of all the resources on the system which could then be used to display all places where documentation is in the system in something like one of the KDE/GNOME help browser, or the info browser...
I wanted to ask the question: what did the move of putting Linux under the GPL do for attracting developers? it would be great to get the response of kernel coders on this. I would speculate that the fact that the kernel belonged to the "community" at large thanks to the GPL and could not be co-opted into another proprietary OS(tcp stacks in NT for example) played a part in assuring developers about the ownership, rightful use, and legacy of their code.
I find RedHat good overall. Havta say I dont like all my software in/usr( eg they mucked around with the KDE setup) for NFS mounting and multiple copies of same software(versions) reasons. I'm ok for most system stuff, but I'm sure most of you remembered hown GNOME rpms mucked wih libgtk, and thus gimp, and how you had to go through a song and dance to get both to co-exist. rpm's which install into/usr/local are so lovable...:-) (or relocatable ones)
My worry is not about Red Hat's motivations, but (a)possibly about market share arrogance..."we own most of the market so our standards are defacto standards and these are the ones we'll use"? (b) with a IPO, they must care about their shareholders, not about us. Will their investor base be savvy enough?
Strangely, gtk/gnome is a better option for commercial concerns. Suppose I am a zip-tools maker, and I want to dock a zip-tools app into the panel. I'm selling my tools commercially. With KDE i now need to pay Troll. With GNOME, I just write it. Probably not a concern for companies, but may be one for shareware authors, small companies, as Qt is 1000+ per developer.
I was wondering, how do they do bootup in a window? Once we get to the rc.sysinit stage (with reference to a RedHat box), some dinky X server could be started, but until that stage, what? Has anyone seen this beast? is it just a xterm with scrolling or some KDE app?
I was wondering, given that FreeBSD is sorta centrally controlled, could it not be that it does nor engender the same sense of community?
Do we not feel that Linux is "my" OS regardless of whether we are developers or consumers. I'd say that a lot of this is due to the lack of centralization and some of it is due to the license, the GPL, which provides a sense of protected ownership. The GPL makes us feel that Linux is OUR intellectual property, it belongs to all of us. Perhaps knowing that their work will not be ripped off (read tcp stack) does motivate some developers.
The lack of centralization makes us feel that we can all make a difference. Its hard to see a tightly controlled source base giving people that notion. While a lot of this is just our notions (surely one can compile KDE for linux without ports, or whatever), our notions and emotions do dictate what we do.
In fact, centralization makes some of us uncomfortable, which is why, for example, some people do not want companies to standardize on RedHat. Finally I do believe that the fanaticism and evangelism, dangerous in its extreme form, is a direct product of this sense of community.
Slashdot Mirror
Surprises me that no-one has figured out the real reason. Its commercial.
India's already trying to take some of the Euro satellite launching business away and this only helps further technology in that direction. Groups of developing nations might want to share satellites which India or China would launch for them. O pay to have their own sent up.
Would be good too for developing spy tech, to spy across noth countries various borders. And there will be missile tech spinoffs too.
Agreed entirely with the author about the malaise in Linux desktopland. Infact, I wrote a not so well proposed article about it almost an year back: Whats Wrong in Linux DesktopLand.
Red Hat has since attempted to do a unifying job, but the question I ask is this: how many of the applications that Red Hat ships will they support? As an end user, why ought I have to deal with figuring out how to install mp3 playing or DVD playing capabilities. Why should I not have good fonts? This is the value add commercial outfits are supposed to provide.
But they cant because they are too busy making too many CD's. Why not, as the author has suggested, pare down the number of applications, and pay a royalty to the author/maintainer of the application for each copy sold? Why must a Linux company follow the same path as a standard company in not renumerating the author? (To be fair, RedHat employs many application or subsystem authors, but why not pay the others a royalty instead. For example, why not pay Ximian for eg a royalty to maintain Evolution to be consistent with RedHat design guidelines?).
The authors suggestion of one toolkit is important too. I applaud Lindows and Lycoris for dumping gnome and making KDE based applications (though I dont applaud everything as root idea..why not use capabilities and gradually eliminate root from most applications). I dont agree with KDE as a toolkit choice as the high licensing cost of Qt screws small developers wishing to develop commercial apps or shareware. The Mac is a thriving desktop platform precisely because of these people, and we need to attract such development if we want to keep the long term viability of Linux..dont forget that windows started out as a poor desktop implementation, and but for linux+bsd's would have largely wiped unix out of small and mid-range installations.
LGPL toolkits are good choices...
Here's one possible plan. Create a new distribution, I like to call it birdbrain because thats all the brain one should need to use it. Elitists not welcome. The basic subsystems are kernel+device, init, basic unix utils, binutils, libsystem(libc, curses, etc), directory services/auth. Thats 6 subsystems..create 6 teams, and assign royalties. Get basic X. Pay royalties. Get basic languages: perl, pythonChoose the basic desktop, say gnome. Get Ximian to package it, and get Ximian redcarpet to distribute it. Pay royalties. Choose no more than 15 gnome apps as part of the basic package..choose teams for each, hopefully including original developers, who are willing to fork, customize to needs of distribution. Needs are for (IMO): browser, instant messager, email/news, news aggregator, editor, wysiwyg html editor, rdesktop, file manager, package manager/installer/redcarpet, media player, pdf reader, terminal. Thats all. Make sure media player can play both DVD's and mp3's. If this requires factoring licensing costs into the distrib, so be it. USABILITY comes FIRST. Then choose personal server apps for fileserving, personal web serving, ssh serving.
Thats it in the basic system. If this sounds like taking a page out of Apple, well, yes it does, except that the whole commercial idea here is to get money directly to the developers who maintain the app for the distribution. Think of it as debian on a much smaller, and thus way more coherent scale.
Now make add on packages with separate royalty and responsibility spheres, for development(compilers), science(plotting, etc), office. Anf of these packages, and also the previous 15 odd apps, ought to replacable by others provided they provide the same task capabilities. Nautilus can then be made more task oriented too, where tasks are done independent of the apps providing them.
Create an experimental distrib in which new things are played with before being dropped into the stable distrib. Examples would include a unified way to deal with data in text form like Apple's plists or RedHat's al
I'm not sure I understand, but
modulo shared libraries, dosent having two loaded GRE's increase the
memory bloat at the expense of disk bloat. Last I knew disk was much
cheaper than memory...
Phoenix extensions would seem to be the only way to go then..but why are phoenix extensions better than standard mozilla ones?
Everybody seems to be missing the larger point here. This capability is the stepping stone to being able to put up spy satellites. In a few years India will be able to keep tabs on Pakistans cross border infiltration and nuclear installations, and hopefully, in the case of fundamentalists taking over( which I dont think will happen, Pakistan being to a large extent a fairly sensible country), be able to make a pre-emptive strike.
If you want to learn more about the origins of this programme read Abdul Kalam's Wings of Fire. Its a very inspiring book. That Kalam is now President of India(which is a titular position without much power, unlike the Prime Minister), is
itself a testament to where self-reliance and competition in science and technology can take one.
Hopefully the programme can now be commercially self sufficient, and the pace of space exploration and missile defence research becomes faster. As you have probably realized in the last year, South Asia ia a tough neighborhood: a dictatorship to the west and east(Pak and Burma), the worlds largest communist state to the North, and ofcourse, central asia and the unstable 'stan's near by..
Lastly, such development can only serve as a long term counterpoise to scary go-it-alonists and US supremacists like some members of this administration...
I work at a university, and I can see clearly who is switching.
Those who say they wont switch here are probably system administrators. Since I do sysadmin as part of my job, I can say that that part of me is a control freak, and loves the power of linux. That is also the reason why Linux has it hard on the desktop: only macosx, lycoris, lindows are even thinking of deprecating root in their OS'es.
The part of me which programs is split. Doing scientific programming today is easier on linux because of the number of high quality numerics/graphics libs available for X11. This will change. However, have you seen the simplicity of macosx? Every app is a directory. No gtk compatability problems(for those who remember). Copy the app anywhere. click, go. For command line people, change defaults using the default command, since all apps use plists. Open any file by saying open bla.pdf. It will use the default app. use open -with if you want a specific app.
The linking model is simple. The loading model is simple. applescript scripts most apps and is way easier to use than COM or bonobo. Still linux is a familiar model to lots of people. So I know people now, grad students and post-docs and engineers, whose desaktop is a macosx box and who program on linux..the professors dont program much so macosx works well for them. This student/scientist/engineer/programmer is the only remaining market.
But at the end of the day its the apps. Excel is available. And itunes and iphoto just rock.
There was a time when i liked struggling with linux to get all this working. At some point, one just wants to code. One dosent want to deal with dependencies, etc. You will say apt-get and I'll say hallelujah, its a great thing, but why cant i just install the freaking app where I want it too, and delete it by trashing it. rpm --erase??? Who would think of that?
The sad part is, most of what macosx has done could and still can be done on linux. Make a restricted distribution. Share earnings with app developers. Choose 10-15 best-of-breed apps, thats all. Thing of the next evolutionary step in these apps, rather than remaining behind the curve. root should only be a single user mode thing. Like gentoo, make init scripts dependent on whats running and whats not. Simplify the runlevels to single-user, and multi-user. Reduce hardware complexity by certifying systems based on linux friendly manufacturers. run daemons not as root. Get rid of the start, or hat, or whatever menu. Get rid of the XP like icons(see redhat8 beta). Give gtk a default look which dosent look like grey shit. Use a tasteful muted color scheme. Make sure pcmcia and usb and firewire just work on plug in. Use hotplug and devfs like mandrake do. Get rid of one million etc config files and use gconf and alchemist like redhat do. Simplify the gnome2.0 desktop. Check out the innovations in oe-one's desktop. Use autofs pervasively. Implement per process namespaces. Implement a simple event layer on top of bonobo, pipes, mimetypes, clipboard, etc to make scripting the desktop trivial. See plan9's plumbing. Unify zsh(bash) and nautilus to use same mime system. Allow apps to be manipulated as directories. When such directories are opened in either, allow hooks to be called which can start or install apps into a dependency database. Create a pasteboard server like in macosx. Implement gnustep over gtk2.0...
You get my point. There is so much thats already there but just missing a bit. It needs people with that extra bit of innovation, and that extra bit of compansation a app-royalty scheme would generate to push it across the edge. It needs that part of me that is a system administrator to let go. But it may be too late.
Exactly why isnt Michael quitting slashdot then, since slashdot is the spawn of the evil proprietary software producing VA.
Sheesh, talk about integrity.
And while we are at it, if a proprietary piece of software with an OPEN FILE FORMAT was used to get me my refund or license or whatever faster, so be it. Let the people who actually do the work choose the best tool for the job.
That link should have been
http://reno.cis.upenn.edu/~rahul/venn.jpg
Combining forces is a step in the right direction. But how can 4 companies survive competing for the same small services market for servers, where you are selling to system admins who know what to do anyway, and will buy only hardware related support? Yeah, some will sell as insurance, but I'd rather buy my insurance from IBM who will be around tomorrow.
.
The only market that would have bought in droves, and did between 1999-2002 is the academic and technical workstation market, where the need was plenty and the expertise thin. Where was the user centric product then? Today's linux companies are making the same mistakes that the unix guys did, leaving the desktop to be picked up by microsoft, concentrating on the server. And today, OSX is replacing linux as the desktop of choice for power users.
And whats with 1000 packages with 10 email clients, all substandard? Why package 2 desktops? Make a courageous decision and pick one! Why duplicate work, decreasing productivity? Do RedHat and Ximian need to both package gnome or evolution? Why not contract it out? Why not pay the gphoto developers a royalty, something you can do if you had just 15 desktop apps? Linux is presently sustaining programmers through VC, not through profits. This isnt a get-rich-quick scheme. Support the developers. Provide user testing for them. Give them a chance to live atleast part time of their software and consulting
But most of all, dont leave the desktop. For they who dont have the desktop today wont have the server tomorrow.
I wrote some more about this stuff, see http://3point0.nareau.com and the 2 links on that page. Also see
http://reno/~rahul/venn.jpg for an example of how an ecology of companies around a linux distribution and an application server(spacestation on desktop and cloudserver on server) could work.
And email me at tig@nareau.com if you want to do something about creating a distribution with one desktop(gnome as gtk is lgpl, i believe in letting developers choose their own license, so no religious nut jobs pls), few well done apps, attention to quality, user interface and simplicity rather than emphasis on service contracts, and a biz-model by which work is distributed to the actual package developers on a per-product sold royalty basis, and selling the software is supposed to bring in money.
I am concerning myself basically with the desktop. Personally, I dont think linux is doing badly on the server now(but see down), though I question the need for 5 not-so-heavily-differentiated distributions.
I wrote an essay on this late last year; its at
http://3pointo.nareau.com/stories/storyReader$5
One question worth asking is after the 3 years of VC 98-2000, why do we still not have a decent set of true type fonts, free.
The point is, the desktop was abandoned by existing companies for the server. Seems to have been a sensible notion by the companies?
NYET! History has shown that dominance on the desktop today leads to a dominance on the server tomorrow. Unix already dropped the ball in the early 90's. We are doing it again.
Lets get this clear first, Nautilus is a great and beautiful product. In my book both are important.
In March 2001, Eazel laid off 40 employees, paring itself down to 35. Even at its height, there were 13 out of 15 developers on Nautilus paid for by Eazel. So there were 60 other people in this company. Ok so estimate 5 execs, 5 graphic designers, 5 marketing and sales(they werent selling anything yet), 3 sysadmins, 7 customer support, 5 administrative, and we get to 30 more people. Obviously at no time did they seem to need to be more than 45 people (though 3,2,2,2,2,2 =13 more = 28 total which is what they got down to seems more realistic).
They got 11 million in March 2000, 1 year at 100,000 per employee for 75 employees ate 7.5 million straight, leaving 3.5 million+founders investments. At 30, they would probably still have had another year to try and validate their business model, or be acquired.
So I get the feeling that their burn rate was too high, too dot-commy, and that lead to their going under. Of course its easier to say this in hindsight, but they probably expected to clinch a second round easy. They should have been warned by last april's crash.
There are now many companies with embedded versions of linux. Most of their web sites have no provisions to give me the source.(Given that really what these companies sell is a config.mk, thats ok I guess). But a lot of these companies make changes I would think to make linux small, include special functionalities. And without providing a listing of whats free and whats proprietary module add-on how am I to know that they are not cheating the GPL(too many me-too companies out there for the nasdaq buck these days..).
Attitudes like Calderas have ensured fractionation. Some of the lack of cohesion in embedded development is tailoring for different purposes I am sure. But see how nicely linux developed all these years, with the older companies(even caldera) playing by the rules. Qualitatively, I get the feeling things have changed.
I'm telling my friends and family that caldera is a bad buy(software and stock). They are a company without a focussed market thrust anyway. First it was openlinux desktop, now it is some e-crap, tomorrow it will be something else. RedHat, Debian, and Mandrake(distribs) and VA always give back their efforts to the community and are not leeches.
I have a parallel port zip drive, would it
be possible to boot linux off that?
If you look at the page carefully, you will find that the money is being put up by the US Government through the National Labs. So the US Govt is now *directly* funding Open Source. The products would be used to further parallel programming, beowulf, and just plain normal programming in a significant way. Yippee!
,papers, for details.
Why Python? A lot of developers at Labs are Biologists, physicists, chemists, weather folk, etc who develop large beouwulf and other numerical codes. Given that most scientists pick up programming without any formal training, they are more interested in spending time on their coding and science than in Makefile and autoconf arcana(and dont tell me that a m4 based config system is nor arcane).
Python is easyly understandable. A java/C++ programmer would pick it up in 2 hours. A C programmer in a day or two. People without any formal training pick it up in a week or so. existing C and C++ and Fortran code can be easily wrapped using SWIG--this has been used to implement the outer calculation and visualization loop for massively parallel molecular dynamics programs. See http://www.swig.org
This is the basic point: How do you extent the advantages of software enginnering to those who are not software developers, but simulators and scientists, and other engineers. And how do you stop being elitist and lower the barrier of entry into Open Source for the nonexperts and the common man. And how do you ensure extensibility. The answer is scripting, and good coding practises for the non-expert demand python.
This was exactly what I thought when I heard of the Sun Ray. I think they made up their own protocol, undocumented, though.
Two ways to do this on Linux:
1)Create Protocol. Use Xvfb virtual framebuffer
server and use protocol to communicate
2)Use xvnc X server, use linux svgalib vnc client, or write a new vnc client for the linux framebuffer. Someone had linux svgalib client on a floppy. Pop into a PC, boot PC, and you have a walking thin-client. Could even store a private key if needed. Replace floppy by card. Done. Tunnel vnc over ssh for encryption and compression and use cards to provide private key and encrypted password sent over network. (Second layer of protection)
Second layer could be, desktop linux PC's with CODA used to hoard releavant parts of linux servers. Could be laptops. VNC clients could still
connect to these machines. Such a laptop would solve the problem of server downtime and use from anywhere in the world where there is no bandwidth.
Companies may not want it(secrets). However, it
could be used for workgroup synchronization and multiple options in case of failure.
Anyone want to do this? Email me..
(rahul@reno.cis.upenn.edu) Suns offering is really
quite simple and we could have a product free in
software and $400 in hardware INCLUDING monitor,
utilizing new PC's, or people could use their old
PC's as NC's. I call it Sting Ray!
The LDP docs, in the old days of Sunsite and few or no RPM's and debs, were crucial to anyone doing
..., so that I can zoom in on exactly what I want. The more general point here is that finer granularity needs to be intelligent
/usr/doc, and stuff is in html there, umm no its in PDF/PS, or oh, in tex and I need to process it....Now
Linux. Times have changed, and the LDP (IMHO) needs to organize documentation btter I think.
Most People do work in the context of a
distribution these days, and distributions do
things differently. I've often found files at
different places than in the docs. The docs maybe
too detailed for me to spend my time on...I want
to code, not administer my systems, nor compile
everything from scratch. What I am saying is that
the granularity of the docs is not fine enough--I'd be happy to know how to use Linuxconf to set up routing with a small sidebar as to what is going on behind the scenes. I dont want to read
the whole Net-3 howto, unless that is my interest.
Finer Granularity implies better structured documentation, in the sense that it is not enough
to say "section", but section on routing with references to the route giving the error message set
granularity, so that searching is smarter, and natrual language searches are guided.
Now Docbook provides a lot of these tags. You can write API and command man-pages, there are tags like command, syopsis, errormsg, funcsynopsis,
screen, screenshot, input(what the user inputs) etc. Some of these tags are common to all documentation, some ought to be removed, and be put into domain specific XML dtd's, like those involving class framework API's, etc
Docbook is too big to use. A whittled down set of tags, finer grained and more functional than linuxdoc, combined with domain specific XML dtd's
which can then be combined with docbook XML using
namespaces ought to be developed. What would have to be agreed upon, or volunteers found to tag, is
a minimal docbook subset, and a method of combining DTD's. Conforming to these specifications, developers would be free to create their own DTD's, or convertable doc formats, and we would all reap the benefits of structured documentation.
Finally, from a users point of view, there is the documentation finding chaos. Umm lets see
do man bla, is that fails, and even if that succeeds for a FSF software, do info bla, umm we're still not there, wait there is
obviously this chaos will remain as everyone has
their favorite formats. But searching is out of the question, and at the very least, we need a RDF
source(like newhoo) of all the resources on the system which could then be used to display all places where documentation is in the system in something like one of the KDE/GNOME help browser, or
the info browser...
I wanted to ask the question: what did the move of putting Linux under the GPL do for attracting developers? it would be great to get the response of kernel coders on this. I would speculate that the fact that the kernel belonged to the "community" at large thanks to the GPL and could not be co-opted into another proprietary OS(tcp stacks in NT for example) played a part in assuring developers about the ownership, rightful use, and legacy of their code.
I find RedHat good overall. Havta say I dont /usr( eg they mucked around with the KDE setup) for NFS mounting /usr/local are so lovable...:-) (or relocatable ones)
like all my software in
and multiple copies of same software(versions) reasons. I'm ok for most system stuff, but I'm sure most of you remembered hown GNOME rpms mucked wih libgtk, and thus gimp, and how you had to go through a song and dance to get both to co-exist.
rpm's which install into
My worry is not about Red Hat's motivations, but (a)possibly about market share arrogance..."we own most of the market so our standards are defacto standards and these are the ones we'll use"?
(b) with a IPO, they must care about their shareholders, not about us. Will their investor base be savvy enough?
Strangely, gtk/gnome is a better option for
commercial concerns. Suppose I am a zip-tools
maker, and I want to dock a zip-tools app into the
panel. I'm selling my tools commercially. With KDE
i now need to pay Troll. With GNOME, I just write
it. Probably not a concern for companies, but
may be one for shareware authors, small companies, as Qt is 1000+ per developer.
I was wondering, how do they do bootup in a window? Once we get to the rc.sysinit stage
(with reference to a RedHat box), some dinky X server could be started, but until that stage,
what? Has anyone seen this beast? is it just a xterm with scrolling or some KDE app?
I was wondering, given that FreeBSD is sorta
centrally controlled, could it not be that it
does nor engender the same sense of community?
Do we not feel that Linux is "my" OS regardless
of whether we are developers or consumers. I'd
say that a lot of this is due to the lack of centralization and some of it is due to the license, the GPL, which provides a sense of
protected ownership. The GPL makes us feel that
Linux is OUR intellectual property, it belongs
to all of us. Perhaps knowing that their work
will not be ripped off (read tcp stack) does
motivate some developers.
The lack of centralization makes us feel that we
can all make a difference. Its hard to see a
tightly controlled source base giving people that notion. While a lot of this is just our notions
(surely one can compile KDE for linux without ports, or whatever), our notions and emotions do dictate what we do.
In fact, centralization makes some of us uncomfortable, which is why, for example, some people do not want companies to standardize on RedHat. Finally I do believe that the fanaticism and evangelism, dangerous in its extreme form, is a direct product of this sense of community.