Does anyone on Slashdot really use UML to document a design?
Only after the product is delivered, and only if the customer really insists. Why? For two reasons.
Firstly, and this is not a criticism of UML but of crap management, because the specs for a project are usually non-existant or subject to continual change. I have become accustomed to working on projects where the "design" is a verbal description of the data sources and a photoshopped mockup of the site. Then the requirements change up until the day before the site goes live. Testing? That word is a bit like "referential integrity" in MySQL - it appears to be understood, but doesn't have any effect.
Secondly, and this is a criticism of UML as well as bad management, UML is seen as a panacea for everything. It's just SSADM for a new generation. The managers who have a vague notion of what UML is see it as a silver bullet. They want UML diagrams, and then use them as a stick to beat you with when ludicrous timescales mean that a product is late. This kind of manager thinks a UML diagram means that the project is as good as written.
What's the answer? Better informed managers, who insist on detailed requirements documents which have been sanity checked by competent programmers before they're signed off. The customer (usually an internal one in another department of the same company) has to be told at some stage that their wilder flights of fancy are not doable.
On my copy of "Pull My Strings" Jello stops the intro to "California Uber Alles" and says "... we're not a punk band we're a new wave band...". I think the point was that the major labels had taken the new wave style and manufactured a radio friendly version with a band called The Knack. That's why the chorus of "Pull My Strings" sounds like The Knack's "My Sharona", albeit with the words changed to "My payola".
Trent Reznor "helped" Filter by falling out with his guitarist on the Broken tour. The guitarist in question was Richard Patrick who then went on to form Filter.
Frontline Assembly were originally Michael Balch and Bill Leeb. Rhys Fulber only joined when Michael Balch left following the "Gashed Senses and Crossfire" album. Personally I only liked one more album they did, "Caustic Grip", which included the jaw dropping "Provision". Then like most of the late eighties industrial bands they disappeared up their own arse and produced watered down techno.
Cool link. I've been getting frustrated lately because loads of tabs are now written in the Guitar Pro format, and I wasn't aware of an open source program that could display them.
Nouvelle Vague are well worth checking out as well. They do Bossanova versions of alternative and new wave classics like "Too Drunk To Fuck" by the Dead Kennedys and "Marian" by the Sisters. The vocals are all done by heavily French accented female singers - on the Dead Kennedys cover the vocalist actually sounds drunk...
You can't be making that good a living if you have to wash you underwear out in the yard. Us Unix programmers can afford to use the slot machines at the laundromat.
The "pointer turns into a caret" idea makes it difficult to distinguish between the mouse pointer and the real caret (insertion point). A transparent or opaque pointer when you pass over a text entry or editor widget sounds like a great idea though.
My understanding of Xen is that NetBSD is one of several operating systems that can run as the primary or "monitor" OS within Xen. Other operating systems (including Windows and Linux) can then be run as secondary OS'es. A research version of Windows could run as the primary / monitor OS, but it is not readily available.
I should add that I've not actually tried Xen yet, but if my understanding of Xen is correct then I'm keen to try it. I'd ideally like to have NetBSD as the primary OS, and Linux as a secondary to test the portability of server side C++ code I'm writing.
So, your Solaris servers haven't been patched in at least two years?
Speaking for my own company, we have big database servers that are usually only patched for stability reasons. Security patches are much lower priority as the machines in question are thoroughly firewalled and only allow access to a handful of ports.
Our webservers are a different matter, and get all security and stability patches. However, they are a stack of much smaller machines behind a load balancer, and we can pull a few of them out of service at any time.
From every person that I know that got the "privilege" to meet RMS in person the expression "smelling nice" usually is not in their description...
A few years ago, RMS was in Oxford, UK for some talks he was doing. His accomodation fell through, so I offered to put him up in my flat for a few days. My abiding memories of his stay are that he drank gallons of herbal tea and took really long showers. In fact he was quite bemused by how small the typical British hot water tank is, as it meant he couldn't shower for as long as he liked...
The IsNot operator described in the patent also differentiates between objects with the same address in different memory spaces.
I'm sure CORBA has been doing this for years with one of the methods in the base CORBA::Object class. I think the method's called equivalent(), but it's been a while since I did any CORBA - in fact I'm having to endure the pathetic MicroSoft alternative that is SOAP.
To reply to both people saying they would compile for Gentoo servers elsewhere, it's still a pain in the arse. I don't have the time to compile my own packages, and I don't see any signed and checksummed packages for Gentoo from sources I can trust. I want to be able to install a set of packages that I am confident have been well tested on a base operating system that doesn't differ from my own. RHEL is a win here, because the slow release cycles are partly down to a thorough QA process. I'm also more confident that when I go to Oracle with a problem, I'm going to get serious consideration whereas Gentoo users are more likely to get the brush off because of the wide scope for tweaking their system.
There's a lot of history behind the Rosegarden developers choice of Qt and KDE. The original Rosegarden dates from the early 1990's, when free toolkits for X GUI development were limited to pretty much nothing but Athena. Rosegarden managed to look good, which was a remarkable acheivement for any application written with the Athena toolkit.
In the late 1990's, the original Rosegarden developers wanted to do a ground up rewrite using a modern GUI toolkit. One of the main developers is very keen on OOP, so C++ was the obvious choice for the implementation. This was attempted using bits of the then nascent GNOME platform, but got bogged down because the GTK+ binding for C++ was not up to scratch.
The dissatisfaction with the GTK+ bindings for C++ lead to the current incarnation of Rosegarden, which uses Qt and KDE. The Rosegarden developers have commented a number of times that using the Qt/KDE framework has saved them having to reinvent the wheel on a number of occasions.
GNOME is my preferred desktop, and GTK+ is my preferred toolkit for GUI development. However, I'm more than happy to install the base KDE libraries to run Rosegarden, as nothing else can touch it for sequencing on Linux.
What does red hat do to make that same kernel so much more stable than kernel.org? If an application is screwing things over, logical step is to drop it.
RedHat does the kind of stress testing using common usage patterns and edge cases, and on a scale that loosely organised volunteers currently don't. I'm not saying that Debian (for example) couldn't come up with a project that does this kind of thing, but this is the area that RHEL appeals to (non-pointy haired) bosses.
The conservative release cycles of RHEL are because the users can't afford the downtime required by the frequency of Fedora upgrades (and Gentoo is a non-starter for enterprise users unless you're a masochist who likes getting ragged on by your boss when the system crawls during an emerge).
Well that's f*cked off the Debian and Gentoo amateurs...
If you want to brag about portability, talk about CPU ISAs (and perhaps MMUs)
Well, perhaps if you want to brag you can talk about CPU ISA's. I'd rather consider cleanliness and portability of the code, which is where NetBSD wins hands down. FreeBSD sacrificed portability for x86 optimisation, OpenBSD just doesn't have the developer base (most of their ports are lifted wholesale from Net) and Linux is just a joke.
Comparing that list to the list of NetBSD ports it is now evident that Linux has been ported to more platforms than NetBSD.
The important thing to note is the past tense - has been ported. Even if it was true that Linux had been ported to more platforms than NetBSD, it is a bit misleading as many of those ports were incomplete, were never satisfactorily merged into the mainstream kernel and have since rotted away. Where Greg KH shows his ignorance is in his confusion of platforms (or architectures) with a single processor type. NetBSD runs on many architectures that use a similar processor type. Wherever possible, userland and device driver code is shared, but this does not mean that my NeXT box is the same as my Macintosh LC II.
The VMS kernel was rewritten in a high level language for the Alpha port (not necessarily C as parts of VMS are written in a number of well known and not so well known languages). At the same time, compliance with POSIX was added and the resulting system became OpenVMS.
As for Cutler's comments on Unix, he is most certainy biased. Anyone reading Bach's "Design of the Unix Operating System", Tannenbaum's "Operating Systems" or the BSD Daemon book has to be impressed with the cleanliness of the Unix design and implementation.
Which is why you should go with Linux, rather than NetBSD, since NetBSD's notion of "support" runs to "we can get it to boot on this hardware" rather than "works well on this hardware and supports all common drivers"
Your description of NetBSD's hardware support is a bit odd. Unlike Linux where the philosophy is "if it works on one platform include it", NetBSD is designed so that a driver written for one platform will work reliably on all those that support the same device. As for missing support for some devices, that's no more or less true than Linux. Compare the support for SGI MIPS for example - when it works in Debian it supports a similar subset of systems and devices. Likewise for the Vax, although here Linux supports a tiny suset of the systems and devices that NetBSD does. Linux has a large proportion of platforms that are only notionally supported (they booted once upon a time, but have languished thanks to Linus being primarily concerned with i386).
all things considered, Gentoo is a better bet
I don't know of a single embedded company using Gentoo (embedded boards being the subject of the original article), and all the embedded Linux kits I've used have been based on RedHat. I also know of no companies using Gentoo on their servers - again it is RedHat along with the odd SuSE box. I did encounter Debian once as a contractor, but the guy who had installed it was reputedly a loose cannon and ahad done it without auhorisation (one of my first tasks was to switch the machine in question to RedHat).
How could you mention Delia Derbyshire without mentioning the rather famous sci-fi TV theme tune she realised from the music by Ron Grainger?
Sadly she later left the BBC, as their was a high degree of male chauvinism in the corporation back then. Initially she had been sneered at for making thhe Doctor Who theme ("too weird"), and then jealousy quicked in when the theme became a revered classic. I can't give you any references for this, bur it comes from a close friend who's an A/V engineer at the Beeb.
I'd be very suprised if Debian actually supported it as well as NetBSD, based on the dubious support for non-i386 architectures. Debian's Sparc support is hit and miss, and their SGI MIPS support is a joke.
Slashdot Mirror
Does anyone on Slashdot really use UML to document a design?
Only after the product is delivered, and only if the customer really insists. Why? For two reasons.
Firstly, and this is not a criticism of UML but of crap management, because the specs for a project are usually non-existant or subject to continual change. I have become accustomed to working on projects where the "design" is a verbal description of the data sources and a photoshopped mockup of the site. Then the requirements change up until the day before the site goes live. Testing? That word is a bit like "referential integrity" in MySQL - it appears to be understood, but doesn't have any effect.
Secondly, and this is a criticism of UML as well as bad management, UML is seen as a panacea for everything. It's just SSADM for a new generation. The managers who have a vague notion of what UML is see it as a silver bullet. They want UML diagrams, and then use them as a stick to beat you with when ludicrous timescales mean that a product is late. This kind of manager thinks a UML diagram means that the project is as good as written.
What's the answer? Better informed managers, who insist on detailed requirements documents which have been sanity checked by competent programmers before they're signed off. The customer (usually an internal one in another department of the same company) has to be told at some stage that their wilder flights of fancy are not doable.
On my copy of "Pull My Strings" Jello stops the intro to "California Uber Alles" and says "... we're not a punk band we're a new wave band ...". I think the point was that the major labels had taken the new wave style and manufactured a radio friendly version with a band called The Knack. That's why the chorus of "Pull My Strings" sounds like The Knack's "My Sharona", albeit with the words changed to "My payola".
Interesting. How did he help out Filter?
Trent Reznor "helped" Filter by falling out with his guitarist on the Broken tour. The guitarist in question was Richard Patrick who then went on to form Filter.
Frontline Assembly were originally Michael Balch and Bill Leeb. Rhys Fulber only joined when Michael Balch left following the "Gashed Senses and Crossfire" album. Personally I only liked one more album they did, "Caustic Grip", which included the jaw dropping "Provision". Then like most of the late eighties industrial bands they disappeared up their own arse and produced watered down techno.
Cool link. I've been getting frustrated lately because loads of tabs are now written in the Guitar Pro format, and I wasn't aware of an open source program that could display them.
Nouvelle Vague are well worth checking out as well. They do Bossanova versions of alternative and new wave classics like "Too Drunk To Fuck" by the Dead Kennedys and "Marian" by the Sisters. The vocals are all done by heavily French accented female singers - on the Dead Kennedys cover the vocalist actually sounds drunk ...
I'm going to go outside now and wash my boxter s.
You can't be making that good a living if you have to wash you underwear out in the yard. Us Unix programmers can afford to use the slot machines at the laundromat.
The "pointer turns into a caret" idea makes it difficult to distinguish between the mouse pointer and the real caret (insertion point). A transparent or opaque pointer when you pass over a text entry or editor widget sounds like a great idea though.
The joke is on Thompson, he reinvented VMS badly.
VMS wasn't to be available for another eight years when Unix was conceived.
NeXTstep could run usefully in less memory than BeOS. Yes. Really.
As an owner of a NeXT turbo slab I can confirm that NeXTstep runs very well on a Motorola 68040 (30Mhz IIRC) and 16Mb of RAM.
My understanding of Xen is that NetBSD is one of several operating systems that can run as the primary or "monitor" OS within Xen. Other operating systems (including Windows and Linux) can then be run as secondary OS'es. A research version of Windows could run as the primary / monitor OS, but it is not readily available.
I should add that I've not actually tried Xen yet, but if my understanding of Xen is correct then I'm keen to try it. I'd ideally like to have NetBSD as the primary OS, and Linux as a secondary to test the portability of server side C++ code I'm writing.
So, your Solaris servers haven't been patched in at least two years?
Speaking for my own company, we have big database servers that are usually only patched for stability reasons. Security patches are much lower priority as the machines in question are thoroughly firewalled and only allow access to a handful of ports.
Our webservers are a different matter, and get all security and stability patches. However, they are a stack of much smaller machines behind a load balancer, and we can pull a few of them out of service at any time.
From every person that I know that got the "privilege" to meet RMS in person the expression "smelling nice" usually is not in their description...
A few years ago, RMS was in Oxford, UK for some talks he was doing. His accomodation fell through, so I offered to put him up in my flat for a few days. My abiding memories of his stay are that he drank gallons of herbal tea and took really long showers. In fact he was quite bemused by how small the typical British hot water tank is, as it meant he couldn't shower for as long as he liked ...
The IsNot operator described in the patent also differentiates between objects with the same address in different memory spaces.
I'm sure CORBA has been doing this for years with one of the methods in the base CORBA::Object class. I think the method's called equivalent(), but it's been a while since I did any CORBA - in fact I'm having to endure the pathetic MicroSoft alternative that is SOAP.
To reply to both people saying they would compile for Gentoo servers elsewhere, it's still a pain in the arse. I don't have the time to compile my own packages, and I don't see any signed and checksummed packages for Gentoo from sources I can trust. I want to be able to install a set of packages that I am confident have been well tested on a base operating system that doesn't differ from my own. RHEL is a win here, because the slow release cycles are partly down to a thorough QA process. I'm also more confident that when I go to Oracle with a problem, I'm going to get serious consideration whereas Gentoo users are more likely to get the brush off because of the wide scope for tweaking their system.
There's a lot of history behind the Rosegarden developers choice of Qt and KDE. The original Rosegarden dates from the early 1990's, when free toolkits for X GUI development were limited to pretty much nothing but Athena. Rosegarden managed to look good, which was a remarkable acheivement for any application written with the Athena toolkit.
In the late 1990's, the original Rosegarden developers wanted to do a ground up rewrite using a modern GUI toolkit. One of the main developers is very keen on OOP, so C++ was the obvious choice for the implementation. This was attempted using bits of the then nascent GNOME platform, but got bogged down because the GTK+ binding for C++ was not up to scratch.
The dissatisfaction with the GTK+ bindings for C++ lead to the current incarnation of Rosegarden, which uses Qt and KDE. The Rosegarden developers have commented a number of times that using the Qt/KDE framework has saved them having to reinvent the wheel on a number of occasions.
GNOME is my preferred desktop, and GTK+ is my preferred toolkit for GUI development. However, I'm more than happy to install the base KDE libraries to run Rosegarden, as nothing else can touch it for sequencing on Linux.
What does red hat do to make that same kernel so much more stable than kernel.org? If an application is screwing things over, logical step is to drop it.
RedHat does the kind of stress testing using common usage patterns and edge cases, and on a scale that loosely organised volunteers currently don't. I'm not saying that Debian (for example) couldn't come up with a project that does this kind of thing, but this is the area that RHEL appeals to (non-pointy haired) bosses.
The conservative release cycles of RHEL are because the users can't afford the downtime required by the frequency of Fedora upgrades (and Gentoo is a non-starter for enterprise users unless you're a masochist who likes getting ragged on by your boss when the system crawls during an emerge).
Well that's f*cked off the Debian and Gentoo amateurs ...
If you want to brag about portability, talk about CPU ISAs (and perhaps MMUs)
Well, perhaps if you want to brag you can talk about CPU ISA's. I'd rather consider cleanliness and portability of the code, which is where NetBSD wins hands down. FreeBSD sacrificed portability for x86 optimisation, OpenBSD just doesn't have the developer base (most of their ports are lifted wholesale from Net) and Linux is just a joke.
To quote Greg KH's posting:
Comparing that list to the list of NetBSD ports it is now evident that Linux has been ported to more platforms than NetBSD.
The important thing to note is the past tense - has been ported. Even if it was true that Linux had been ported to more platforms than NetBSD, it is a bit misleading as many of those ports were incomplete, were never satisfactorily merged into the mainstream kernel and have since rotted away. Where Greg KH shows his ignorance is in his confusion of platforms (or architectures) with a single processor type. NetBSD runs on many architectures that use a similar processor type. Wherever possible, userland and device driver code is shared, but this does not mean that my NeXT box is the same as my Macintosh LC II.
The VMS kernel was rewritten in a high level language for the Alpha port (not necessarily C as parts of VMS are written in a number of well known and not so well known languages). At the same time, compliance with POSIX was added and the resulting system became OpenVMS.
As for Cutler's comments on Unix, he is most certainy biased. Anyone reading Bach's "Design of the Unix Operating System", Tannenbaum's "Operating Systems" or the BSD Daemon book has to be impressed with the cleanliness of the Unix design and implementation.
Well, according to the description of the following image on the screenshots page it runs on MacOS X:
Bluefish running on a Mac, in french.
Code Veronica was set on a remote island and an antarctic ice station.
Which is why you should go with Linux, rather than NetBSD, since NetBSD's notion of "support" runs to "we can get it to boot on this hardware" rather than "works well on this hardware and supports all common drivers"
Your description of NetBSD's hardware support is a bit odd. Unlike Linux where the philosophy is "if it works on one platform include it", NetBSD is designed so that a driver written for one platform will work reliably on all those that support the same device. As for missing support for some devices, that's no more or less true than Linux. Compare the support for SGI MIPS for example - when it works in Debian it supports a similar subset of systems and devices. Likewise for the Vax, although here Linux supports a tiny suset of the systems and devices that NetBSD does. Linux has a large proportion of platforms that are only notionally supported (they booted once upon a time, but have languished thanks to Linus being primarily concerned with i386).
all things considered, Gentoo is a better bet
I don't know of a single embedded company using Gentoo (embedded boards being the subject of the original article), and all the embedded Linux kits I've used have been based on RedHat. I also know of no companies using Gentoo on their servers - again it is RedHat along with the odd SuSE box. I did encounter Debian once as a contractor, but the guy who had installed it was reputedly a loose cannon and ahad done it without auhorisation (one of my first tasks was to switch the machine in question to RedHat).
How could you mention Delia Derbyshire without mentioning the rather famous sci-fi TV theme tune she realised from the music by Ron Grainger?
Sadly she later left the BBC, as their was a high degree of male chauvinism in the corporation back then. Initially she had been sneered at for making thhe Doctor Who theme ("too weird"), and then jealousy quicked in when the theme became a revered classic. I can't give you any references for this, bur it comes from a close friend who's an A/V engineer at the Beeb.
I'd be very suprised if Debian actually supported it as well as NetBSD, based on the dubious support for non-i386 architectures. Debian's Sparc support is hit and miss, and their SGI MIPS support is a joke.