Yeah, I had GConf in mind when I wrote the post, actually (although a more generalised version would be nice). XML's more-complex syntax is made up for with its expressive power, in my opinion, but that's just me. I just feel that XML would be a better choice since it can natively support more complex data structures than a first-order mapping. I know a lot of *nix hackers like to be able to directly manipulate their configurations, but you have to look at it from the angle that this is still possible -- most text editors provide automatic XML completion and so on, it's basically a matter of balancing act between usefulness and complexity, I suppose.
I choose XML because it's an expressive, general, standardised format. I'm not sure another proprietary format is really what we need at this point. Even a fairly general app (curses, command line, or GUI versions aren't too hard to conjure up in the mind) which abstracts the process would ease editing. I know people like to have "direct control", and you'll never really lose that, but some programs need (and many already use) the extra expressive power that XML can provide, and a general framework should probably encapsulate the most complex examples as well as the simpler ones.
I do really want to hear people's opinions on this though, it's something I've thought about for a while, especially in the light of GConf.
Additionally, it could be interpreted much more logically by a secondary program which would give a consistent, whole-system view of configuration. We have all these metaphors for data representation, what say we use them? Many systems already do.
Additionally, such a system is not a "single point of failure". Redundant storage (which you would probably want for most configuration files anyway) and the structured nature of XML lends itself to repair.
A lot of people need to get over the fact that text-editing files is probably not the best way to configure a system. Besides, looking at many more-modern systems, they already use XML for configuration.
I'd say using a large, registry-like XML database might be more sensible. Could get unmanagable in size, I guess, but something like RDF would probably fix that.:)
Further to the sibling post, Mozillazine's Extensions Dev page has a wealth of fantastic resources for creating stuff. Once you get into the nitty-gritty, XULPlanet is mighty handy (and probably constitutes a lot of the "documentation" you require. Also, O'Reilly's Mozilla book is available free online.
I dunno. It seems like needless FUD to single out.NET as a "bad thing". The only real complaint is the size of the software, and it's not huge on most systems built in the last 6 or so years.
I understand that it could cause problems, but I'm fed up with the cavalier "It's installing MS software! That's worse than the fact it's breaking my config, monitoring my net usage, and sending my bank account details to some shady character in northern Africa!" attitude that's being thrown around. Most types of install that it could incite could potentially cause a compatibility problem of some kind, but at least this is software that will be harmless on most systems.
"Pointless" was perhaps a bit strong, but it still strikes me as a tremendous non-issue. Perhaps a humourous oversight on the part of the malware writers, but little else.
The pointless complaint is that it's installing.NET specifically. Most loaders like this install something without the user's permission. Keyloggers, viruses, trojan horses, whatever. Yet the article singles out the.NET framework. A non-harmful API. It might be more sizable (not by a significant amount) but it's a damn sight less harmful than most things that malware choose to install.
Likewise Linux, OS/2, or whatever other general-purpose OS you feel like shoving onto an ATM. The fact is that if your system is on a secure network of some kind (which, if it isn't a legal requirement, should be a legal requirement), your system should be isolated from outside influence in any respect that matters here.
As for the Spyware thing, it's trivially as easy to bundle a (single user) Adware product along with a Linux software package (and most other systems) as with a Windows one. I agree with your assertion that there's been problems in the past with some Windows products, but I don't think Windows Embedded has been any worse hit than any other comparable product. When a system has a limited use it is obviously far easier to protect it from malicious code.
(I'm not trying to troll, and I'm interested in hearing your thoughts about this.)
Never thought you were! It seems unfortunate you'd start with such an assumption of attitudes. Just the internet I guess...
I'd reply, but the sibling post beginning " On another point, HTML and TCP/IP are HEAVILY stress tested." answers your criticisms far better than I could see me doing myself.:)
It seems unlikely that an ATM would be designed, or allowed, to run code which was not provided by those in charge of the ATM. The rendering engine is not a problem in this instance -- you're rendering code you've written yourself. These are not web browsers we're talking about, the application is much thinner.
Interesting. I've used M2 in the past, but could never get the thing to work! It just infuriated me for some reason. I'd give it a second look, but it's integrated into their browser which I don't use (unless there's a standalone version -- I'll look into it).
Has there been patches that break the system with Windows Server 2003? I know that a lot of people reckon XPSP2 "broke" Windows, but to be fair there wasn't any software which didn't have an update, and they did recommend you update (and procided a list of software to update), so I look upon people who have problems with SP2 with great skepticism in most cases. I know nothing about Windows Server (I've used it for a total of about 10 minutes, ever) though.
Back to the topic at hand, another post mentioned that Ars Technica switched to Windows in order to use.NET as their platform for web services, which is interesting. It reminds me of a discussion which was being had about higher-level languages vs. lower-level languages here just the other day -- when does extreme efficiency/stability* get outweighed by reduction in development time, and lack of bugs in the specific software for the system?
* Obviously "stability" doesn't really apply to C or low-level languages, but in this case you could be construed to be losing stability by the forced use of a Windows platform. Still not completely convinced of that myself, though.
You know, I prefer GMail's "labels" metaphor for mail to the "folders" metaphor used by most mail clients (and IMAP). I find it easier to organise and find my emails through the GMail interface than through Thunderbird.
As a (probably pointless) piece of evidence to the contrary, just to try and put some faith in the 2.6 series, the system I'm using now has never been able to boot any 2.4-based distro I've tried on it, and 2.6 works flawlessly. I realise this is the equivalent of "happy events in the news" and will probably be ignored, but I try to help:)
That said, the system is currently running Windows XP (hey, I might desire free software, but I need games)...
I'd like to read it if they did -- it would be nice to see technically-minded people explaining the benefits of a system like.NET, even if only to cause the detractors of the system here and elsewhere hunker down and give a more reasoned argument. There's time's it's not given the credit it deserves, it is (in general) one of MS's actual successes.
This is probably going to sound patronising (and it's not intended like that), but surely it's true that stability isn't the only thing affecting uptime. Windows Server 2003 still needs to be restarted on some patches -- that's not a stability flaw, it's a design flaw affecting uptime. Also, Linux can (apparently) handle a higher network load than a comparable Windows machine -- that's not stability, that's resistance to network pummelling. And if you're basing uptime stats on reachability, that'll affect it too.
I'm not saying you're wrong, mind, but the numbers you quote aren't the be-all and end-all of stability.
Wow, what a troll. Of *course* the kernel is device dependent. It's a fricking kernel.
No, your parent post wasn't really a troll -- more a criticism of the Linux kernel's monolithic structure. It's an "academic vs real world" argument, really. I'd swither towards supporting a microkernel system, but the fact is, Linux is here and it works (most of the time), whereas microkernel systems are restricted to the HURD, which is gonna work "real soon now", and commercial offerings from Apple and Microsoft.
Empirical evidence has shown, time and time again, that low level languages like C and Forth produce more efficient,
Granted, but with the way hardware is going it's not really a big factor most of the time.
faster,
This is the same thing.
and easier to maintain code than today's so called high level languages.
It seems unlikely there's empirical evidence to prove this, primarily because it's unlikely to be true in any fair test. Higher-level languages are quicker to develop in and less likely to have fatal bugs -- these, in most production environment, are by far the most important factors.
Unless you see a speed increase of Olog(n) then you are simply wasting resources by using a high level language. The reason for this is simple, the amount of time spent in coding is finite, yet, oddly enough, CPU time isn't. Thusly unless you can improve an algorithm along that order then your time was wasted. This is simple first year college stuff here, hardly rocket science.
If your algorithm isn't improved, you wasted time writing it in a lower-level language. Why? It took longer to write. 99% of the time, the efficiency you gain will be completely outweighed by the extra time you spent developing the system. Too much emphasis is placed on "efficiency" by many, when it's simply not needed.
Obviously, though, this is something that can be debated until you're blue in the face. But discrediting higher-level languages (I assume you mean Java, Python, Perl, C#, whatever) when there is clear empirical evidence of their benefits, simply because you do not see their benefits as benefits, doesn't seem right.
As for the "college profs" comment, most that I've met won't "throw stuff at you". Hell, the programming languages guy I have likes C, and argues for its use where efficiency is needed, to the detriment of development time and code safeness. Many professors like functional languages, and it's hard to blame them, given the terseness of the languages and their uses in reasoning. But, in general, they gain little from "pimping the new stuff". At least here.
Same here, with certain exceptions (which are mentioned in the sibling posts of this one). The TV licensing people can obtain a warrant through a court if you refuse them entry, although they generally don't. They only visit houses of people who have no registered TV license, and if you send them a note saying you have no TV they usually leave it be, with possibly a single check...
Slashdot Mirror
Yeah, I had GConf in mind when I wrote the post, actually (although a more generalised version would be nice). XML's more-complex syntax is made up for with its expressive power, in my opinion, but that's just me. I just feel that XML would be a better choice since it can natively support more complex data structures than a first-order mapping. I know a lot of *nix hackers like to be able to directly manipulate their configurations, but you have to look at it from the angle that this is still possible -- most text editors provide automatic XML completion and so on, it's basically a matter of balancing act between usefulness and complexity, I suppose.
I choose XML because it's an expressive, general, standardised format. I'm not sure another proprietary format is really what we need at this point. Even a fairly general app (curses, command line, or GUI versions aren't too hard to conjure up in the mind) which abstracts the process would ease editing. I know people like to have "direct control", and you'll never really lose that, but some programs need (and many already use) the extra expressive power that XML can provide, and a general framework should probably encapsulate the most complex examples as well as the simpler ones.
I do really want to hear people's opinions on this though, it's something I've thought about for a while, especially in the light of GConf.
XML is designed to be human-readable.
Additionally, it could be interpreted much more logically by a secondary program which would give a consistent, whole-system view of configuration. We have all these metaphors for data representation, what say we use them? Many systems already do.
Additionally, such a system is not a "single point of failure". Redundant storage (which you would probably want for most configuration files anyway) and the structured nature of XML lends itself to repair.
A lot of people need to get over the fact that text-editing files is probably not the best way to configure a system. Besides, looking at many more-modern systems, they already use XML for configuration.
I'd say using a large, registry-like XML database might be more sensible. Could get unmanagable in size, I guess, but something like RDF would probably fix that. :)
But oh no! They've countered with Bittorrent! More traffic makes it work better!
There's torrent links in the article (on the names of the formats).
Further to the sibling post, Mozillazine's Extensions Dev page has a wealth of fantastic resources for creating stuff. Once you get into the nitty-gritty, XULPlanet is mighty handy (and probably constitutes a lot of the "documentation" you require. Also, O'Reilly's Mozilla book is available free online.
I dunno. It seems like needless FUD to single out .NET as a "bad thing". The only real complaint is the size of the software, and it's not huge on most systems built in the last 6 or so years.
I understand that it could cause problems, but I'm fed up with the cavalier "It's installing MS software! That's worse than the fact it's breaking my config, monitoring my net usage, and sending my bank account details to some shady character in northern Africa!" attitude that's being thrown around. Most types of install that it could incite could potentially cause a compatibility problem of some kind, but at least this is software that will be harmless on most systems.
"Pointless" was perhaps a bit strong, but it still strikes me as a tremendous non-issue. Perhaps a humourous oversight on the part of the malware writers, but little else.
The pointless complaint is that it's installing .NET specifically. Most loaders like this install something without the user's permission. Keyloggers, viruses, trojan horses, whatever. Yet the article singles out the .NET framework. A non-harmful API. It might be more sizable (not by a significant amount) but it's a damn sight less harmful than most things that malware choose to install.
Likewise Linux, OS/2, or whatever other general-purpose OS you feel like shoving onto an ATM. The fact is that if your system is on a secure network of some kind (which, if it isn't a legal requirement, should be a legal requirement), your system should be isolated from outside influence in any respect that matters here.
As for the Spyware thing, it's trivially as easy to bundle a (single user) Adware product along with a Linux software package (and most other systems) as with a Windows one. I agree with your assertion that there's been problems in the past with some Windows products, but I don't think Windows Embedded has been any worse hit than any other comparable product. When a system has a limited use it is obviously far easier to protect it from malicious code.
I'd reply, but the sibling post beginning " On another point, HTML and TCP/IP are HEAVILY stress tested." answers your criticisms far better than I could see me doing myself.
It seems unlikely that an ATM would be designed, or allowed, to run code which was not provided by those in charge of the ATM. The rendering engine is not a problem in this instance -- you're rendering code you've written yourself. These are not web browsers we're talking about, the application is much thinner.
The fact is, that 23MB is non-harmful software which will probably help you run other software. It's a pointless complaint, really.
Interesting. I've used M2 in the past, but could never get the thing to work! It just infuriated me for some reason. I'd give it a second look, but it's integrated into their browser which I don't use (unless there's a standalone version -- I'll look into it).
Nice one, now if they would hurry up with a Windows version...
Has there been patches that break the system with Windows Server 2003? I know that a lot of people reckon XPSP2 "broke" Windows, but to be fair there wasn't any software which didn't have an update, and they did recommend you update (and procided a list of software to update), so I look upon people who have problems with SP2 with great skepticism in most cases. I know nothing about Windows Server (I've used it for a total of about 10 minutes, ever) though.
.NET as their platform for web services, which is interesting. It reminds me of a discussion which was being had about higher-level languages vs. lower-level languages here just the other day -- when does extreme efficiency/stability* get outweighed by reduction in development time, and lack of bugs in the specific software for the system?
Back to the topic at hand, another post mentioned that Ars Technica switched to Windows in order to use
* Obviously "stability" doesn't really apply to C or low-level languages, but in this case you could be construed to be losing stability by the forced use of a Windows platform. Still not completely convinced of that myself, though.
You know, I prefer GMail's "labels" metaphor for mail to the "folders" metaphor used by most mail clients (and IMAP). I find it easier to organise and find my emails through the GMail interface than through Thunderbird.
As a (probably pointless) piece of evidence to the contrary, just to try and put some faith in the 2.6 series, the system I'm using now has never been able to boot any 2.4-based distro I've tried on it, and 2.6 works flawlessly. I realise this is the equivalent of "happy events in the news" and will probably be ignored, but I try to help :)
That said, the system is currently running Windows XP (hey, I might desire free software, but I need games)...
I'd like to read it if they did -- it would be nice to see technically-minded people explaining the benefits of a system like .NET, even if only to cause the detractors of the system here and elsewhere hunker down and give a more reasoned argument. There's time's it's not given the credit it deserves, it is (in general) one of MS's actual successes.
This is probably going to sound patronising (and it's not intended like that), but surely it's true that stability isn't the only thing affecting uptime. Windows Server 2003 still needs to be restarted on some patches -- that's not a stability flaw, it's a design flaw affecting uptime. Also, Linux can (apparently) handle a higher network load than a comparable Windows machine -- that's not stability, that's resistance to network pummelling. And if you're basing uptime stats on reachability, that'll affect it too.
I'm not saying you're wrong, mind, but the numbers you quote aren't the be-all and end-all of stability.
It's a real pity, in any case. It's one of the biggest problems with Linux today, although it's obviously been getting better of late.
Obviously, though, this is something that can be debated until you're blue in the face. But discrediting higher-level languages (I assume you mean Java, Python, Perl, C#, whatever) when there is clear empirical evidence of their benefits, simply because you do not see their benefits as benefits, doesn't seem right.
As for the "college profs" comment, most that I've met won't "throw stuff at you". Hell, the programming languages guy I have likes C, and argues for its use where efficiency is needed, to the detriment of development time and code safeness. Many professors like functional languages, and it's hard to blame them, given the terseness of the languages and their uses in reasoning. But, in general, they gain little from "pimping the new stuff". At least here.
Same here, with certain exceptions (which are mentioned in the sibling posts of this one). The TV licensing people can obtain a warrant through a court if you refuse them entry, although they generally don't. They only visit houses of people who have no registered TV license, and if you send them a note saying you have no TV they usually leave it be, with possibly a single check...