>
My big problem with the Hurd is that there is no attempt to provide real-time support
Early optmisation is the root of all sort of evil...
Seriously, I am in no position to know, but wouldn't the multiserver, microkernel design allow for a later realtime version? At least a little bit easier and (or) better than with Linux? And wouldn't focus on microkernel now delay it even more for the server and the desktop?
>
MkLinux (Linux running on the original series of PowerMacs) was a Linux-on-Mach system. In comparison to the monolithic PPCLinux, it had a performance penalty of around 15-20%.
AFAIR, it is still around in order to support NuBus Macs, and the performance penalty is from 10% to 15%.
>
What are the relative benefits of L4 vs the Mach Microkernel?
L4 is being maintained and developed, while GNU would have to maintain Mach all by itself. Also, L4 is much simpler, thus at least potentially more reliable, safe, and a better platform for the multi-server microkernel architecture of the Hurd.
>
As I understand it, MacOS X's microkernel is also based on the Mach microkernel...
Yes and no. I never found anyone who could justify Mac OS X (NeXTStep)'s use of Mach. They take the Mach codebase, but merge it with the BSD kernel; in this way they both loose BSD's simplicity, leanness and performance, and Mach's flexibility. The worse of both worlds.
The only benefit I could gather is the use of drivers made for MkLinux they formerly supported, but then they forfeit the much bigger codebase of, say, Linux or simple BSD drivers, or even the future codebase of L4 drivers, and get stuck where the Hurd is in the Mach codebase: doing it all by themselves as more and more projects get off Mach unto L4, simple BSD or Linux or anything else.
You don't know what you're talking about. They didn't. It is just OO-like functions you access in plain C.
>
C++ is one of the most popular languages out there
So popular everyone spends loads of time trying to fix it, and it grows and grows...
>
restarting flame wars about programming language and license that were settled years ago because both sides realized that they were stupid is a bit weak.
I criticised the claim you need C++ to reuse code, and that's clearly stupid. I called it arcane, and it is. And Debian never settled the KDE issue, it is just that Trolltech recanted and relicensed Qt under the GNU GPL.
I know the guy's role is sociological, not technical, but I get worried about how much confusion reigns in free software.
One can reuse code in any language. It is source code availability that enables that, not C++; even OO in general seems to conduce to reuse but frequently leads to problems such as the weak base objects.
C++, with its huge complexity, is a handicap -- Java and C# are still complex but at least they try to be a little bit simpler, not to mention definetly better things such as functional programming.
Now if KDE wants to keep using an arcane language, it is their choice, just as they insisted for how long in a confusing licensing and now keep insisting in a confusing interface. But don't spread technical misinformation, please.
>
There has been at least one working communist system that was inherently democratic... the communal communes of Spain in the early 20th century
The communes in Spain weren't actually communes in the communistic sense. They are just democratically-managed counties, as you still can find in most of Europe and North America. They do have some property in common -- in England the commons, in Spain the irrigation system and so on -- but most land is still private, people own their homes or lease them... they are more moshavim then kibbutzim, and probably a lot less than moshavim.
The Left has a way of idealising its heroes and demonising its enemies, just like everyone else... actually Washington hardly had any interest against the Spanish communes, being then in isolationism. Actually Washington's interests only crossed the Atlantic when forced to (the Zimmerman telegram, Pearl Harbor) until it was faced with the task of countering Sovietic expansionism. What did happen is that the new, Republican Spanish government was too Leftist for most of the population, being anticlerical and such, and this granted popular support for the Monarchists.
The name of the OS is GNU/Linux. OK, so MS could put Linux, the kernel, under the MS Windows interface and Win32 API -- but what would this buy them? Besides the huge headache of making it work (Win32 is hugely more complex than Carbon, né the Macintosh Toolbox, ever was), they would either compromise Linux or slow MS Windows, as they would loose all types of dirty tricks that get them performance at the cost of stability.
But this wouldn't be the worse. The worse would be getting the world to realise that the problem isn't the kernel, but the API. Actually the MS WNT kernel is quite good, but they have to keep an absurd API to keep application compatibility; Linux without the GNU C library, utils and the X Window System would buy them nothing here.
Now they could adopt the GNU C library and the X Window System, but then Win32 would become just a legacy personality of MS WNT remember it already support a half-GNU and an OS/2 ones. Integrating old Win32 apps in the new X environment would be a huge headache, and then they would have in effect just a better WINE...
And anyway, this would do little for them... OpenOffice.org is already most there, v2 or v3 should prove on par with MS Office. A straightforward Debian GNU/Linux Gnome environment would still perform better, be more stable, and eventually simpler than any kind of WinX contraption.
This is sure good for standards-compliance, but it still won't do any difference for sites that rely on Java or Flash, which still effectively block many or all non-x86 users, specially Flash.
I still remember when Intergraph had the nicest technical workstations you could ask for without going RISC... too bad they really fooled themselves into thinking you can work with Wintel and yet survive on quality.
>
If your RDBMS is doing a full table scan just to do a count(*)
COUNT (*) is an SQL construct, and SQL isn't relational, therefore an RDBMS can't do COUNT (*) unless it implements a quasi-SQL backward compatibility mode (such as Alphora Dataphor's RealSQL).
>
Apple's Remote Desktop system actually installs and uses PostgreSQL for all its data storage
Does it do that in a standard way so that other users can take advantage of that same PostgreSQL installation, or if one wants a general-purpose database one still needs to do a new installation?
Anyway, a PostgreSQL installation is so light that it hardly makes any difference...
The term it relates to is "open systems," which refers to standardized Unix systems, as opposed to mainframes.
This is completely wrong.
This days you can consider a mainframe open, because not only it can ran GNU/Linux but also it has a facility to run open systems programs, protocols etc. Other non-Unix systems do so, like Digital VMS.
Open systems are systems that implement open standards, that is, standards agreed upon by representative bodies like ISO specifying interfaces, protocols, file formats etc.
It fazes me... why not write AppleWorks documents, do Apple expect everyone to migrate at once? Perhaps they plan to start doing that later, maybe when they get also a spreadsheet and a DB frontend?
As it is really less than AppleWorks, MS shouldn't care... but I guess Apple is only too happy that OOo exists to take care of MS Office users should need arise.
>
We've got Linux servers in the server room, which we ssh to.
No need even for ssh, you can have diskless X terminals booting from the same host it will open an X session into, or a terminal with X in the firmware that uses XDMCP to find the most available host.
>
Microcomputing is all about not having to share with others.
Problem is, except for gaming, 3D animations, video and the like, sharing is good and not sharing is dumb.
For a fraction of the price of full-blown PCs for each person in a house, building or block, you can have one big, mean server and X terminals or Sun Rays in each room. Usually the CPU and bus is idle waiting for users' input or peripherals, so you not only get a fraction of the cost but also the full speed of a multiprocessing machine with fast buses. Finally, you get a professionally managed server, and with free software all the programs you could need with no virus, no worries about installation and licensing...
>
I haven't got a clue on how shared libraries work across multiple users. Do they map at all? If so, it should be easy to load it at boot as 'nobody', and when the real user logs in, the OS will see "Hey, it's already loaded, let's just point a reference".
No need for all that. Even MS knows how to load a common code segment for all programs, and having each one have its own data segment. What MS couldn't do until at least MS WNT was how to do that for multiple concurrent users (time sharing), but that has been common in Un*x for ages now.
To sum it up, the first user who launches a program gets all the libraries loaded, all the others use the libraries already there.
The big benefit of a separate Gecko and Necko installation would be the possibility of having Mozilla and Firefox using the same installation of them, and having other Gecko browsers such as Epiphany or Galeon depend on smaller packages.
Would you please stop confounding everyone out of the US with your euphemisms? It's leftism, not liberalism. Liberalism relates to liberty, not to government intervention. It means economical laissez-faire and parlamentary democracy.
Slashdot Mirror
Early optmisation is the root of all sort of evil...
Seriously, I am in no position to know, but wouldn't the multiserver, microkernel design allow for a later realtime version? At least a little bit easier and (or) better than with Linux? And wouldn't focus on microkernel now delay it even more for the server and the desktop?
So why can't you say the other parts of the Hurd are the right way to do things? Specifically, please -- pointers would suffice.
AFAIR, it is still around in order to support NuBus Macs, and the performance penalty is from 10% to 15%.
L4 is being maintained and developed, while GNU would have to maintain Mach all by itself. Also, L4 is much simpler, thus at least potentially more reliable, safe, and a better platform for the multi-server microkernel architecture of the Hurd.
Yes and no. I never found anyone who could justify Mac OS X (NeXTStep)'s use of Mach. They take the Mach codebase, but merge it with the BSD kernel; in this way they both loose BSD's simplicity, leanness and performance, and Mach's flexibility. The worse of both worlds.
The only benefit I could gather is the use of drivers made for MkLinux they formerly supported, but then they forfeit the much bigger codebase of, say, Linux or simple BSD drivers, or even the future codebase of L4 drivers, and get stuck where the Hurd is in the Mach codebase: doing it all by themselves as more and more projects get off Mach unto L4, simple BSD or Linux or anything else.
You don't know what you're talking about. They didn't. It is just OO-like functions you access in plain C.
So popular everyone spends loads of time trying to fix it, and it grows and grows...
I criticised the claim you need C++ to reuse code, and that's clearly stupid. I called it arcane, and it is. And Debian never settled the KDE issue, it is just that Trolltech recanted and relicensed Qt under the GNU GPL.
That is it, they couldn't. It took years to make Trolltech change Qt's license to something GNU GPL-compatible.
And suffering, see how progress slowed down and Linus found himself unable to produce stable kernels in 2.6.
But yes, C is a lot less arcane than C++
Gnome is plain C. And yes, Objective C is quite less arcane than C++
No, Trolltech Qt's license wasn't compatible with the GNU GPL. Go inform yourself before replying.
Perhaps not being a native English speaker I misused the word -- I meant a language you need to be a language lawyer to be effective in.
See why Debian was unable to distribute KDE...
I know the guy's role is sociological, not technical, but I get worried about how much confusion reigns in free software.
One can reuse code in any language. It is source code availability that enables that, not C++; even OO in general seems to conduce to reuse but frequently leads to problems such as the weak base objects.
C++, with its huge complexity, is a handicap -- Java and C# are still complex but at least they try to be a little bit simpler, not to mention definetly better things such as functional programming.
Now if KDE wants to keep using an arcane language, it is their choice, just as they insisted for how long in a confusing licensing and now keep insisting in a confusing interface. But don't spread technical misinformation, please.
The communes in Spain weren't actually communes in the communistic sense. They are just democratically-managed counties, as you still can find in most of Europe and North America. They do have some property in common -- in England the commons, in Spain the irrigation system and so on -- but most land is still private, people own their homes or lease them... they are more moshavim then kibbutzim, and probably a lot less than moshavim.
The Left has a way of idealising its heroes and demonising its enemies, just like everyone else... actually Washington hardly had any interest against the Spanish communes, being then in isolationism. Actually Washington's interests only crossed the Atlantic when forced to (the Zimmerman telegram, Pearl Harbor) until it was faced with the task of countering Sovietic expansionism. What did happen is that the new, Republican Spanish government was too Leftist for most of the population, being anticlerical and such, and this granted popular support for the Monarchists.
The name of the OS is GNU/Linux. OK, so MS could put Linux, the kernel, under the MS Windows interface and Win32 API -- but what would this buy them? Besides the huge headache of making it work (Win32 is hugely more complex than Carbon, né the Macintosh Toolbox, ever was), they would either compromise Linux or slow MS Windows, as they would loose all types of dirty tricks that get them performance at the cost of stability.
But this wouldn't be the worse. The worse would be getting the world to realise that the problem isn't the kernel, but the API. Actually the MS WNT kernel is quite good, but they have to keep an absurd API to keep application compatibility; Linux without the GNU C library, utils and the X Window System would buy them nothing here.
Now they could adopt the GNU C library and the X Window System, but then Win32 would become just a legacy personality of MS WNT remember it already support a half-GNU and an OS/2 ones. Integrating old Win32 apps in the new X environment would be a huge headache, and then they would have in effect just a better WINE...
And anyway, this would do little for them... OpenOffice.org is already most there, v2 or v3 should prove on par with MS Office. A straightforward Debian GNU/Linux Gnome environment would still perform better, be more stable, and eventually simpler than any kind of WinX contraption.
This is sure good for standards-compliance, but it still won't do any difference for sites that rely on Java or Flash, which still effectively block many or all non-x86 users, specially Flash.
I still remember when Intergraph had the nicest technical workstations you could ask for without going RISC... too bad they really fooled themselves into thinking you can work with Wintel and yet survive on quality.
How long will we survive now that our iniquity (Gn XV:16) is filling up so quickly?
COUNT (*) is an SQL construct, and SQL isn't relational, therefore an RDBMS can't do COUNT (*) unless it implements a quasi-SQL backward compatibility mode (such as Alphora Dataphor's RealSQL).
Does it do that in a standard way so that other users can take advantage of that same PostgreSQL installation, or if one wants a general-purpose database one still needs to do a new installation?
Anyway, a PostgreSQL installation is so light that it hardly makes any difference...
This is not true. StarOffice did have a Mac OS port until at least v4.
This is completely wrong.
This days you can consider a mainframe open, because not only it can ran GNU/Linux but also it has a facility to run open systems programs, protocols etc. Other non-Unix systems do so, like Digital VMS.
Open systems are systems that implement open standards, that is, standards agreed upon by representative bodies like ISO specifying interfaces, protocols, file formats etc.
It fazes me... why not write AppleWorks documents, do Apple expect everyone to migrate at once? Perhaps they plan to start doing that later, maybe when they get also a spreadsheet and a DB frontend?
As it is really less than AppleWorks, MS shouldn't care... but I guess Apple is only too happy that OOo exists to take care of MS Office users should need arise.
No need even for ssh, you can have diskless X terminals booting from the same host it will open an X session into, or a terminal with X in the firmware that uses XDMCP to find the most available host.
Problem is, except for gaming, 3D animations, video and the like, sharing is good and not sharing is dumb.
For a fraction of the price of full-blown PCs for each person in a house, building or block, you can have one big, mean server and X terminals or Sun Rays in each room. Usually the CPU and bus is idle waiting for users' input or peripherals, so you not only get a fraction of the cost but also the full speed of a multiprocessing machine with fast buses. Finally, you get a professionally managed server, and with free software all the programs you could need with no virus, no worries about installation and licensing...
Yep. Or forget about prelaunching and just separate the GRE in a package common to Mozilla, Firefox, Camino, Epiphany, Galeon...
No need for all that. Even MS knows how to load a common code segment for all programs, and having each one have its own data segment. What MS couldn't do until at least MS WNT was how to do that for multiple concurrent users (time sharing), but that has been common in Un*x for ages now.
To sum it up, the first user who launches a program gets all the libraries loaded, all the others use the libraries already there.
The big benefit of a separate Gecko and Necko installation would be the possibility of having Mozilla and Firefox using the same installation of them, and having other Gecko browsers such as Epiphany or Galeon depend on smaller packages.
Would you please stop confounding everyone out of the US with your euphemisms? It's leftism, not liberalism. Liberalism relates to liberty, not to government intervention. It means economical laissez-faire and parlamentary democracy.