Re:The coolest things about QNX
on
The Rise Of QNX
·
· Score: 1
Boy I would not try comparing Linux et al. kernel modules to a truely modular architecture. The interface of the kernel modules makes them incredibly fragile, requiring recompilation with every new minor version of the kernel, even when certain kernel options have been activated. QNX services have a relationship to the kernel much like *nix demons do: sure, they use the kernel, but if you change the kernel around a bit, you don't need new demons, unless you want to employ some new kernel feature.
Frankly, the module system in FreeBSD, Linux, and others is more of a convenient way of keeping portions of the kernel out of memory unless you have a use for it. QNX's big advantage here is the layer of abstraction between these services and the microkernel, using their extremely efficient messaging systems.
And for the record, if you wanted to make a Linux app fit on a floppy, modules would/not/ be the way to go. You should build all the modules you will need into a custom kernel, then place a statically linked executable next to it.
Well, one of the primary arguments for augmented assignment was actually related to the p3k project, and optimization. The reasoning goes, that it is faster to perform self.attribute += 1 than it is to perform self.attribute = self.attribute + 1 because there is one less attribute lookup.
And while this may appear trivial, it becomes somewhat more important once you realize that Metaobject programming in Python often involves overriding the methods for getting and setting attributes with new methods. Every get/set counts.
I don't think we're about to see a flood of perl-like syntax sugar rotting the teeth out of our Python just yet. =)
*laugh* p3k is a blue sky python with type inference, and other functionality. Its a moving target, so to speak, so it doesn't have a specific version number..
What task did you need to go at in a roundabout way? The only one that comes to mind is avoiding circular references, and that one is pretty easy to manage.
Am rather behind RedHat.. It's not my personal preference for Unix, but I do appreciate their dedication in keeping their relationship with their customers as open as possible.
Kudos to Mr. Young for such a mature response, relatively unadulterated by Marketing-speak.
Actually, all it means is that the open source community found less bugs than the closed source community did. Since one is a professional QA department, the other is an aggregation of geniuses, sk1pt k1dd13s and flamethrowers, all of whom may or may not bother to write bug reports.
You're comparing two completely different test environments and their productivity, here. I also HIGHLY doubt RH7.0 is 30 million lines of code, total, and that the bugs reported include most of the known bugs for each individual package.
Counter-FUD is just as bad as FUD. Please, don't indulge in it. You make the rest of us look bad.
Not everything in an species is genetic. Behavior patterns are often acquired from a parent. Who will these new guar emulate? Who will teach them how to survive? Instinct will only carry complex animals so far, especially carnivores.
I doubt that cloning will bring back extinct species in a viable way. A species sitting in a zoo, unable to care for itself, is still extinct. It just doesn't know it, yet.
I take a little umbrage at the whole Open Source movement being considered a 'new' movement. Once upon a time, when microcomputers were the domain of hobbyists, software was generally freely distributed. Albeit the source form was assembly, but it was there, and it was free for the manipulating.
Somewhere along the lines, it got profitable, and the money grubbing capitalists got involved. The Open Source movement, per se, is an attempt to break back out of that all-for-me kind of software design, and recapture the attention of mainstream computer users.
So, I don't think anyone, outside of our patron acronyms, RMS and ESR, could really have any kind of claim to 'starting the open source revolution.'
Well.. I don't mind seeing these drops. While I, personally, follow the nightlies with zealous fervor, I understand that many casual users aren't willing to go through the hoops to install PSM for SSL capabilities (Granted, now that XPI is working correctly, PSM is just a click away), or a Java VM, etc. Also, they've turned off a lot of the debugging, so at least the Windows users are going to see the true speed of this thing. It's outperforming IE5 on my poor widdle 266MMX.
Netscape's Previews tend to include the entire kitchen sink. True, you have to rip out that wretched netphone, and most of us would rather have our toenails removed with pliers than use AOL for messaging, but it gives you a good picture of how Moz is coming along as a full application.
There are a couple unique issues that I've found with this build, which I've already reported, but so far, it looks quite good.
Well, that's the entire point. To Romanic and Teutonic language speakers, the postfix seems a little odd.. But, in Korean, where much of the context is denoted using suffixes, it might make more sense.
Frankly, I've done so much Forth, I think frontways, backways and sideways.
Pity the poor parser! "Hmmm.. Is it a number? Is it a variable?" We'd be back to using prefixes to specify whether a token is an identifier or a 'number'.
Grrrr.. That'll teach me not to preview slashdot's plain-old-text entries. Bad slash, bad! Evidently, putting <some-tag-here> in/plain text/ causes slashcode to omit it.
Anyway, that last expression was <arg>,,, <verb> <arg>,,,
I believe he is more referring to the alteration of the form, relative to various languages, than just converting the tokens. An example, would be Korean, which could be thought of as a natural language RPN, where the general syntax is:
While there are a few special cases, the overall language is extremely precise, and while fairly position independant when considering object position, specifies that the verb always concludes the clause.
A programming language stemming from a Korean-oriented mind might result in a language with a function call like such:
( "The time is: %t":fmt, [()time]:data )fmtstring;
-which is equivalent to-
( [()time]:data, "The time is: %t":fmt)fmtstring;
As opposed to English, which is relatively position-dependant in syntax:
fmtstring( "The time is: %t", [time()] )
Of course, it's interesting to note that many procedural languages have adopted the (,,, ) convention, which I assume stems from the difficulty in parsing,,,,,,;)
Maybe it's a honey pot, and maybe he'd like to see what you try to attack it. I've done similar, in other forums, of course, and stuck an IDS system next to the target just to see what came by.
Most of it was boring stuff, but I saw a couple interesting tactics.
Yes, it's a threat. There are at least two Orbit-relevant gnome vunerabilities listed on Bugtraq, search for keyword 'orbit', and several more associated with gnome.
I don't care whether they are security conscious or not. I don't want my word processor listening to a port, and I/definately/ don't want that port to be one accessible by an outside influence. Hell, I don't even want the guy two offices down to see it.
That's right.. And noone using nmap might have noticed that friendly little tcp port out there by his lonesome, or an entry in the services, or looked at netstat or anything..
Well, in case you were worried, Slashdot wasn't the first to mention these vunerabilities. There have been vulns listed all over Bugtraq about Gnome. Orbit being wrapped in tcp_wrappers was a fair to middling belt-and-suspenders solution, but it's still going out the door with dangerous default settings, and tcp_wrappers is not sufficient protection on a multiuser machine by itself.
Whoa! Careful.. Blocking the SYNs will stop most trivial and well formed attack, but there are still FIN scans and other attacks that don't have the SYN flag set.
There are a whole raft of exploits involving the use of forged ip fragments that would get around that rule. While your average script kiddie doesn't seem to be using them, it's only a matter of time until someone automates it for those chimps.
*cough* That's assuming there are no problems in your TCP/IP stack. Which is much like assuming the stock market is going to go up.. Usually you're right.. But boy, when you're wrong...;)
I believe it's a legacy problem stemming from the fact that ORBit is a cornerstone of gnome, and being Corba, is capable of remote messaging with other hosts for various savory, and unsavory reasons.
HelixCode probably didn't notice that the default config was wide open, at first. Whups.
Actually, about that 486 claim.. I had an IBM 701c Thinkpad that didn't handle a 10/100mbps 3com Ethernet under RH6.1 very well in a 100-only environment. It would just fall all over itself with big downloads.
Mind, you're probably not going to get that with Road Runner, but one never can tell. =)
There is always the concern that one defense or nother will have a hole. I personally would advocate using a different OS for your firewall, as well, in this case OpenBSD and ipfilter is my favorite, than you use for your interior workstations.
Configure ipfilter in a nice and paranoid fashion, ensure via nmap that no services are addressable, (inetd, portmap, etc, etc.. )/then/ go about placing ipchains input rules on your interior boxen.
The idea is, even if there is a vunerability found in the firewall box, there is a different type of wall protecting your secured hosts. ipfilter is also available for FreeBSD, in case you find OpenBSD a little too user/hardware-hostile.
It's a trivial motion to install a new sound player. Now, for a windows user to install a new web browser, it'll take him a while, and it might not be effort worth taking.
Also, the initial push from MP3 to OGG will come from home users, who are maintaining personal libraries of music, and want better/tighter/faster compression. I don't know if OGG will necessarily win inside a year, but it's definately on the right track to beating MP3's market share.
Slashdot Mirror
Boy I would not try comparing Linux et al. kernel modules to a truely modular architecture. The interface of the kernel modules makes them incredibly fragile, requiring recompilation with every new minor version of the kernel, even when certain kernel options have been activated. QNX services have a relationship to the kernel much like *nix demons do: sure, they use the kernel, but if you change the kernel around a bit, you don't need new demons, unless you want to employ some new kernel feature.
/not/ be the way to go. You should build all the modules you will need into a custom kernel, then place a statically linked executable next to it.
Frankly, the module system in FreeBSD, Linux, and others is more of a convenient way of keeping portions of the kernel out of memory unless you have a use for it. QNX's big advantage here is the layer of abstraction between these services and the microkernel, using their extremely efficient messaging systems.
And for the record, if you wanted to make a Linux app fit on a floppy, modules would
Well, one of the primary arguments for augmented assignment was actually related to the p3k project, and optimization. The reasoning goes, that it is faster to perform self.attribute += 1 than it is to perform self.attribute = self.attribute + 1 because there is one less attribute lookup.
And while this may appear trivial, it becomes somewhat more important once you realize that Metaobject programming in Python often involves overriding the methods for getting and setting attributes with new methods. Every get/set counts.
I don't think we're about to see a flood of perl-like syntax sugar rotting the teeth out of our Python just yet. =)
*laugh* p3k is a blue sky python with type inference, and other functionality. Its a moving target, so to speak, so it doesn't have a specific version number..
Yet.
Actually, I was asking for specifics, because I'm in the mood to give Distutils a workout and write an extension. =)
What task did you need to go at in a roundabout way? The only one that comes to mind is avoiding circular references, and that one is pretty easy to manage.
Am rather behind RedHat.. It's not my personal preference for Unix, but I do appreciate their dedication in keeping their relationship with their customers as open as possible.
Kudos to Mr. Young for such a mature response, relatively unadulterated by Marketing-speak.
Actually, all it means is that the open source community found less bugs than the closed source community did. Since one is a professional QA department, the other is an aggregation of geniuses, sk1pt k1dd13s and flamethrowers, all of whom may or may not bother to write bug reports.
You're comparing two completely different test environments and their productivity, here. I also HIGHLY doubt RH7.0 is 30 million lines of code, total, and that the bugs reported include most of the known bugs for each individual package.
Counter-FUD is just as bad as FUD. Please, don't indulge in it. You make the rest of us look bad.
Not everything in an species is genetic. Behavior patterns are often acquired from a parent. Who will these new guar emulate? Who will teach them how to survive? Instinct will only carry complex animals so far, especially carnivores.
I doubt that cloning will bring back extinct species in a viable way. A species sitting in a zoo, unable to care for itself, is still extinct. It just doesn't know it, yet.
I take a little umbrage at the whole Open Source movement being considered a 'new' movement. Once upon a time, when microcomputers were the domain of hobbyists, software was generally freely distributed. Albeit the source form was assembly, but it was there, and it was free for the manipulating.
Somewhere along the lines, it got profitable, and the money grubbing capitalists got involved. The Open Source movement, per se, is an attempt to break back out of that all-for-me kind of software design, and recapture the attention of mainstream computer users.
So, I don't think anyone, outside of our patron acronyms, RMS and ESR, could really have any kind of claim to 'starting the open source revolution.'
The CTO of Red Hat claims to have invented the internet, with cofounder and Vice President Al Gore.
Well.. I don't mind seeing these drops. While I, personally, follow the nightlies with zealous fervor, I understand that many casual users aren't willing to go through the hoops to install PSM for SSL capabilities (Granted, now that XPI is working correctly, PSM is just a click away), or a Java VM, etc. Also, they've turned off a lot of the debugging, so at least the Windows users are going to see the true speed of this thing. It's outperforming IE5 on my poor widdle 266MMX. Netscape's Previews tend to include the entire kitchen sink. True, you have to rip out that wretched netphone, and most of us would rather have our toenails removed with pliers than use AOL for messaging, but it gives you a good picture of how Moz is coming along as a full application. There are a couple unique issues that I've found with this build, which I've already reported, but so far, it looks quite good.
'coz hell is freezing over. Potato is out! ;)
Well, that's the entire point. To Romanic and Teutonic language speakers, the postfix seems a little odd.. But, in Korean, where much of the context is denoted using suffixes, it might make more sense.
Frankly, I've done so much Forth, I think frontways, backways and sideways.
Pity the poor parser! "Hmmm.. Is it a number? Is it a variable?" We'd be back to using prefixes to specify whether a token is an identifier or a 'number'.
Grrrr.. That'll teach me not to preview slashdot's plain-old-text entries. Bad slash, bad! Evidently, putting <some-tag-here> in /plain text/ causes slashcode to omit it.
;)
Anyway, that last expression was <arg>,,, <verb> <arg>,,,
Time to submit a feature-fix.
I believe he is more referring to the alteration of the form, relative to various languages, than just converting the tokens. An example, would be Korean, which could be thought of as a natural language RPN, where the general syntax is:
,,, ) convention, which I assume stems from the difficulty in parsing ,,, ,,, ;)
clause -> [object].. verb [tense]
sentence -> [object].. [clause conjunction].. clause term
While there are a few special cases, the overall language is extremely precise, and while fairly position independant when considering object position, specifies that the verb always concludes the clause.
A programming language stemming from a Korean-oriented mind might result in a language with a function call like such:
( "The time is: %t":fmt, [()time]:data )fmtstring;
-which is equivalent to-
( [()time]:data, "The time is: %t":fmt)fmtstring;
As opposed to English, which is relatively position-dependant in syntax:
fmtstring( "The time is: %t", [time()] )
Of course, it's interesting to note that many procedural languages have adopted the (
Maybe it's a honey pot, and maybe he'd like to see what you try to attack it. I've done similar, in other forums, of course, and stuck an IDS system next to the target just to see what came by.
Most of it was boring stuff, but I saw a couple interesting tactics.
Yes, it's a threat. There are at least two Orbit-relevant gnome vunerabilities listed on Bugtraq, search for keyword 'orbit', and several more associated with gnome.
/definately/ don't want that port to be one accessible by an outside influence. Hell, I don't even want the guy two offices down to see it.
I don't care whether they are security conscious or not. I don't want my word processor listening to a port, and I
That's right.. And noone using nmap might have noticed that friendly little tcp port out there by his lonesome, or an entry in the services, or looked at netstat or anything..
Well, in case you were worried, Slashdot wasn't the first to mention these vunerabilities. There have been vulns listed all over Bugtraq about Gnome. Orbit being wrapped in tcp_wrappers was a fair to middling belt-and-suspenders solution, but it's still going out the door with dangerous default settings, and tcp_wrappers is not sufficient protection on a multiuser machine by itself.
Whoa! Careful.. Blocking the SYNs will stop most trivial and well formed attack, but there are still FIN scans and other attacks that don't have the SYN flag set.
There are a whole raft of exploits involving the use of forged ip fragments that would get around that rule. While your average script kiddie doesn't seem to be using them, it's only a matter of time until someone automates it for those chimps.
*cough* That's assuming there are no problems in your TCP/IP stack. Which is much like assuming the stock market is going to go up.. Usually you're right.. But boy, when you're wrong... ;)
I believe it's a legacy problem stemming from the fact that ORBit is a cornerstone of gnome, and being Corba, is capable of remote messaging with other hosts for various savory, and unsavory reasons.
HelixCode probably didn't notice that the default config was wide open, at first. Whups.
Actually, about that 486 claim.. I had an IBM 701c Thinkpad that didn't handle a 10/100mbps 3com Ethernet under RH6.1 very well in a 100-only environment. It would just fall all over itself with big downloads.
Mind, you're probably not going to get that with Road Runner, but one never can tell. =)
There is always the concern that one defense or nother will have a hole. I personally would advocate using a different OS for your firewall, as well, in this case OpenBSD and ipfilter is my favorite, than you use for your interior workstations.
Configure ipfilter in a nice and paranoid fashion, ensure via nmap that no services are addressable, (inetd, portmap, etc, etc.. ) /then/ go about placing ipchains input rules on your interior boxen.
The idea is, even if there is a vunerability found in the firewall box, there is a different type of wall protecting your secured hosts. ipfilter is also available for FreeBSD, in case you find OpenBSD a little too user/hardware-hostile.
For more info on ipfilter visit The IPFilter HOWTO and The IPFilter Mailing-Lists.
(Zealotry Notice, *BSD-fanatic who wouldn't trust ipchains with Bill Gate's homepage, but understands that Helix is addictive.)
It's a trivial motion to install a new sound player. Now, for a windows user to install a new web browser, it'll take him a while, and it might not be effort worth taking.
Also, the initial push from MP3 to OGG will come from home users, who are maintaining personal libraries of music, and want better/tighter/faster compression. I don't know if OGG will necessarily win inside a year, but it's definately on the right track to beating MP3's market share.