Re:Always nice to see
on
.NET at JavaONE
·
· Score: 3, Interesting
I agree. When was the last time you saw Bill Gates stand up in front of an audience (microsoft company meetings aside) and say that McNealy or Ellison and their companies are a bunch of mindless jerks that'll be the first against the wall when the revolution comes (or some other childish insult).
The funny thing is that it's not Microsoft that's killing Sun in the marketplace right now, it's open source projects like *BSD and Linux (with a little help from IBM).
But it just wouldn't be fashionable for McNealy to admit that they're losing money to free software (which they are), that's something that only big-bad-bill (or maybe ballmer - he's a litle bigger) would say. On the other hand it's much more acceptable to point the ($1e9) finger of blame at Microsoft. Hey, everyone's doing it!
Does he seriously thing that if Microsoft disappeared off the face of the planet that everyone would suddenly buy Sun? Yeah right. You could either double your cost on hardware/software by going from Microsoft to Sun/Oracle/iPlanet or do it for free. Sure you may have a slightly higher TCO with open source, but have you seen what an Oracle DBA charges (I mean... Oracle's just sooo easy to use:-} )?
but he's not forced to work there is he? surely he can just go somewhere else to find a job. there's a huge difference between cheap labor and slave labor - ask a slave.
Unfortunately, like NS4.7, Opera is really behind on the standards front. Web developers are happy that IE and Mozilla (and derivatives) now both support a considerable chunk of the standards. I've had requests to support opera on the sites that I've done, and I've just had to reply "sorry, choose another browser".
Check Opera's own specs for a shameful list of omissions, specifically the woefully undersupported DOM.
opera may be nice, but its rendering engine supports only a fraction of the w3c spec whereas gecko and mshtml are both usefully close. in fact, I'd say that its scriptable DOM support is worse than NS4.7.
actually C# (or to be more specific CIL) is a little bit easier to compiler to native code than java. The types of operands on the JVM stack are encoded in the bytecode instructions (as opposed to explicitly in the metadata for CIL) which means you have to do more complex dataflow analysis.
The entire microsoft platform SDK toolset is freely available for download on their site. Even if you already have Visual Studio (which you don't need), I'd recommend getting this is it has the latest headers and tools. It comes complete with compilers, libraries, documentation, tools, debuggers, etc... All it's lacking is the nice UI.
The.NET Framework is also freely available for download. Again, it comes with everything you need to build.NET applications, except the nice UI (use vim/emacs/sharpdevelop...)
hang on, is this the same Sun that Microsoft paid about $50M to for a Java license, agreed to do (read: fix the ridiculous bugs and inefficiencies in) the reference implementation on Windows for and then paid another $10M to to get them to shut up when they whined about some of the improvements they made?
Wasteful practices like buying competitors
Hah. Mergers & Aquisitions by large companies are things that small technology companies have wet dreams about. It's an investment of capital. Growth. How do you think that companies like AT&T, Shell, Exxon, Disney, AOL, MCI got so big? It's not like MS has a monopoly on M&A.
Supply and demand says cost of software should be zero.
You know nothing of such things. Supply and demand in a free market states that the value of something is equivalent to what people are willing to pay for it: the cost. It costs money to develop software (people have to eat), if people didn't think it was worth paying for, they wouldn't pay for it. Would you be prepared to work for free? You could argue that MS has been found guilty of having a monopoly, but that's only in the desktop: people have been paying for software of all kinds for a long time. Why? Because it's worth something to them. That's what money is: a number associated with value. Again, simple economics.
The people who write it would rather you use if for free and improve it.
Sure, people who don't want to spend money developing it want you to help them. Because they're not getting any return on their investment, so they can't invest in costly developers - they have to get free ones. But it's not 1960 any more, you don't have to have a PhD to use a computer. In fact, everybody's using them now and 95% of those people don't know, and couldn't give a damn, about how they work, and they'd be completely incapable of fixing stuff even if they wanted to (which they don't). You do, probably, and are able to, possibly, you're an enginee, right? But most people see the value in software not because they can tinker with it, but beacuse they can use it to do some other task that they couldn't do as easily without it.
That's where the value is, and that's where the cost comes from: they're willing to pay for the convenience. Great so there's free software. But most of it is written by people that have other jobs that bring in the bacon (eg Linus/Transmeta) and the value of that software as percieved by the majority of computer users isn't as high as that of professional software. Sure it's popular, but that's mainly because it's free and something that's free is always very compatitively prices, whatever its percieved value.
what, you expect microsoft to spend four years and hundreds of millions of dollars of R&D to produce a product and then just create a new market in which they can reap no benifit from this investment?
Are you fucking crazy?
same old story? sure, this is what every company in existance does: invest, develop, collect. If all your company did was invest and develop then you'd see your cash dry up in a hurry and your investors leaving south even quicker.
It's simple economics, why does nobody on slashdot seem to understand these things? Maybe 'cos the only economics they've ever had to deal with involve getting stuff for free... tanstaafl.
I find the macros quite egregious. It's like programming in some bastardized language.
Oh come on, it's just like programming in C++. Are you trying to tell me you've never used macros before? Ever called getc()? It's a macro. assert()? Macro.
If VC++ supported it (or if you use a compiler that does, like intel's) then you could write a template library that uses partial specialization and expression templates to do the same thing. Would that make you happy? Would it be any different?
And advocating programming by wizard? Come on. That's like admitting its a PITA, and saying "look, here is a tool to make it less of a pain."
What, like memory management? Is memory management really that much of a pain in the ass? We have tools, like garbage collected languages and smart pointers, that do it for you, but does that mean it's such a bad thing? We have RAD development tools like VB, Delphi, Glade, etc... that are essentially glorified wizards, hell even perl hides an extremely complicated and expressive programming environment behind a few keystrokes, it's just a tool. Should we write everything in assembler?
You seem to be contradicting yourself. On one hand you say that things should be simpler, but on the other you say that using tools to make those things easier is bad. What are you trying to say?
COM is a big hack. I'd hate to do all my programming in COM for a distributed system.
Well, I can't argue with your second point (although I wonder if you've actually tried it), but saying that COM is a hack is shortsighted. It's a binary specification for defining reference-counted interfaces between components. An implementation may well be a hack, although I think that ATL is quite elegant. You could argue that reference counting has it's shortcomings, and it has, but it's lightweight and the alternatives are, well, not. C++ has built-in support for dynamic typing (RTTI) but STL instead uses traits templates for passing around type information because it's much more efficient. Is that a hack? Maybe if you could state exactly what it is about COM that you think is a hack, I might understand where you're coming from.
Sure, a command line will give you full control, but a sufficiently rich GUI on top of the same functionality provides a much better learning tool until you're experienced. The functionality for much of windows (Application Center included) is provided by COM controls (WMI, etc...), the cmdline UI and the GUI are both just wrappers for these. Application Center provides a set of command line tools that do the same stuff as the GUI, but since I was new to that particular application I just used the GUI.
(FYI: I've been using command lines since I started programming about 20 years ago)
I'd disagree, it's hardly a pain. There are many ways to do it, you could use VB, Java (with Microsoft's extensions), C# (.NET), C++ (using ATL or MFC) or you could just do it yourself in C.
I prefer ATL since it's flexible and lightweight. All you have to do is define your interface in IDL, and let ATL to do the work:
struct CMyObject :
public CComObjectRoot,
public CComCoClass<CMyObject, &CLSID_MyObject>,
public IMyInterface
{
BEGIN_COM_MAP(CMyObject)
COM_INTERFACE_ENTRY(IMyInterface)
END_COM_MAP()
HRESULT MyMethod () {/*do stuff here...*/ }
};
Hardly a PITA, in fact the ATL COM wizard in VC++ will generate this code for you. Apart from the actual code of your method, everything you need can be done through simple wizards.
COM+ objects running in an Application Center component-load-balanced cluster need to know none of these things. Application Center provides heartbeats based on customizable response functions, virtual IP config and MAC munging.
As I said, all you have to do is write the components and the client application, COM+/AC handles the rest.
I thought I'd just add to this that most people here seem to think that Microsoft's only solution to clustering is the (formerly known as) 'wolfpack' technology.
The Application Center stuff is new (it was supposed to ship with win2k, but was delayed until recently) and although the documentation suggests it's designed for multi-tier web-farms, it'll work fine in a Beowulf-style scenario.
The best feature, for me, was the seamless integration with the COM+ architecture: you write and call your components as if they were running on the same machine (in the same process, even) and the method calls are automatically redirected to the least-loaded cluster member. You can take advantage of object pooling and JIT-activation to reduce the object creation and lifetime-management overheads. I much prefer the object-oriented approach to having bunches of scripts 'rsh'ing all over the place.
Also, the monitoring tools were great for development and testing, but since I didn't use it in a production invironment I can't really attest to their strength under stress.
Here's a link to the Product Guide which has some good screenshots of the UI. That's another thing, I didn't have to read too much to get my cluster working, I just ran through the wizards and tweaked a few options in the dialogs and I was ready to deploy.
check out Application Center. it provides configuration and monitoring for a cluster of windows boxes. also, almost everything on win2k is remotely scriptable through WMI.
I don't see why a command line is even necessary. Why not use RPC or something similar?
While beta testing Application Center I created a cluster of COM+ components, deployed them with 1-click and called them from my client application as if it were a normal method call. I didn't have to write any scripts.
Failover clustering. This is an OS service that servers like SQL Server and Exchange plug into that allows Active/Passive or Active/Active clustering over a shared SCSI/Fibre bus. In theory you could write your app to use this service but I think it would be overkill.
Network Load Balancing. This is just a software version of the standard kinds of NLB found in cisco boxes.
Component Load Balancing. This is the most suitable. It's provided by Application Center and it allows you to deploy COM+ objects on a cluster of machines and have the calls distributed according to the load on those machines. You can control the threading and lifetime of the objects and view the status of the machines pretty easily using the Application Center MMC plugin (or SNMP, I believe). You'd have to wrap the computational part of your application into one or more COM objects. Once you've done that then you can create and call those objects in the cluster as if it were one machine - the clustering is transparent to the client application. I played around with AC a bit when it was in beta for a project that I was working on. We didn't go with it in the end because the design of our application ended up not requiring it (we just went with hardware load balancing), but it seemed like pretty cool technology - if you're into the whole COM thing. It has a really cool rolling deployment feature where you can redeploy your components (and/or IIS application if you have one) to your cluster incrementally while it's still running.
Slashdot Mirror
The funny thing is that it's not Microsoft that's killing Sun in the marketplace right now, it's open source projects like *BSD and Linux (with a little help from IBM).
But it just wouldn't be fashionable for McNealy to admit that they're losing money to free software (which they are), that's something that only big-bad-bill (or maybe ballmer - he's a litle bigger) would say. On the other hand it's much more acceptable to point the ($1e9) finger of blame at Microsoft. Hey, everyone's doing it!
Does he seriously thing that if Microsoft disappeared off the face of the planet that everyone would suddenly buy Sun? Yeah right. You could either double your cost on hardware/software by going from Microsoft to Sun/Oracle/iPlanet or do it for free. Sure you may have a slightly higher TCO with open source, but have you seen what an Oracle DBA charges (I mean... Oracle's just sooo easy to use :-} )?
but he's not forced to work there is he? surely he can just go somewhere else to find a job. there's a huge difference between cheap labor and slave labor - ask a slave.
Check Opera's own specs for a shameful list of omissions, specifically the woefully undersupported DOM.
opera may be nice, but its rendering engine supports only a fraction of the w3c spec whereas gecko and mshtml are both usefully close. in fact, I'd say that its scriptable DOM support is worse than NS4.7.
actually C# (or to be more specific CIL) is a little bit easier to compiler to native code than java. The types of operands on the JVM stack are encoded in the bytecode instructions (as opposed to explicitly in the metadata for CIL) which means you have to do more complex dataflow analysis.
- Use microsoft tools on windows to compile the mono compiler into a
.NET exe.
- Run that
.NET exe (mcs) under mint on linux using the mono class libraries to generate another version of mcs.
Unfortunately the linux-built version of mcs doesn't quite run yet, but I believe it's close to dogfood.windbg, which comes with the platform SDK, supports kernel- and user-mode debugging.
The .NET Framework is also freely available for download. Again, it comes with everything you need to build .NET applications, except the nice UI (use vim/emacs/sharpdevelop...)
The root for the SDK downloads is here
That's where the value is, and that's where the cost comes from: they're willing to pay for the convenience. Great so there's free software. But most of it is written by people that have other jobs that bring in the bacon (eg Linus/Transmeta) and the value of that software as percieved by the majority of computer users isn't as high as that of professional software. Sure it's popular, but that's mainly because it's free and something that's free is always very compatitively prices, whatever its percieved value.
yup, either way they win: fewer people buying Sun boxes...
Are you fucking crazy?
same old story? sure, this is what every company in existance does: invest, develop, collect. If all your company did was invest and develop then you'd see your cash dry up in a hurry and your investors leaving south even quicker.
It's simple economics, why does nobody on slashdot seem to understand these things? Maybe 'cos the only economics they've ever had to deal with involve getting stuff for free... tanstaafl.
please.
check out the recent anime version of Metropilis. It's on limited release, but well worth it if you can make it.
I know there was only one CG shot in the film, was that it?
If VC++ supported it (or if you use a compiler that does, like intel's) then you could write a template library that uses partial specialization and expression templates to do the same thing. Would that make you happy? Would it be any different?
What, like memory management? Is memory management really that much of a pain in the ass? We have tools, like garbage collected languages and smart pointers, that do it for you, but does that mean it's such a bad thing? We have RAD development tools like VB, Delphi, Glade, etc... that are essentially glorified wizards, hell even perl hides an extremely complicated and expressive programming environment behind a few keystrokes, it's just a tool. Should we write everything in assembler?You seem to be contradicting yourself. On one hand you say that things should be simpler, but on the other you say that using tools to make those things easier is bad. What are you trying to say?
Well, I can't argue with your second point (although I wonder if you've actually tried it), but saying that COM is a hack is shortsighted. It's a binary specification for defining reference-counted interfaces between components. An implementation may well be a hack, although I think that ATL is quite elegant. You could argue that reference counting has it's shortcomings, and it has, but it's lightweight and the alternatives are, well, not. C++ has built-in support for dynamic typing (RTTI) but STL instead uses traits templates for passing around type information because it's much more efficient. Is that a hack? Maybe if you could state exactly what it is about COM that you think is a hack, I might understand where you're coming from.(FYI: I've been using command lines since I started programming about 20 years ago)
I prefer ATL since it's flexible and lightweight. All you have to do is define your interface in IDL, and let ATL to do the work:
Hardly a PITA, in fact the ATL COM wizard in VC++ will generate this code for you. Apart from the actual code of your method, everything you need can be done through simple wizards.As I said, all you have to do is write the components and the client application, COM+/AC handles the rest.
The Application Center stuff is new (it was supposed to ship with win2k, but was delayed until recently) and although the documentation suggests it's designed for multi-tier web-farms, it'll work fine in a Beowulf-style scenario.
The best feature, for me, was the seamless integration with the COM+ architecture: you write and call your components as if they were running on the same machine (in the same process, even) and the method calls are automatically redirected to the least-loaded cluster member. You can take advantage of object pooling and JIT-activation to reduce the object creation and lifetime-management overheads. I much prefer the object-oriented approach to having bunches of scripts 'rsh'ing all over the place.
Also, the monitoring tools were great for development and testing, but since I didn't use it in a production invironment I can't really attest to their strength under stress.
Here's a link to the Product Guide which has some good screenshots of the UI. That's another thing, I didn't have to read too much to get my cluster working, I just ran through the wizards and tweaked a few options in the dialogs and I was ready to deploy.
check out Application Center. it provides configuration and monitoring for a cluster of windows boxes. also, almost everything on win2k is remotely scriptable through WMI.
While beta testing Application Center I created a cluster of COM+ components, deployed them with 1-click and called them from my client application as if it were a normal method call. I didn't have to write any scripts.
-
Failover clustering. This is an OS service that servers like SQL Server and Exchange plug into that allows Active/Passive or Active/Active clustering over a shared SCSI/Fibre bus. In theory you could write your app to use this service but I think it would be overkill.
-
Network Load Balancing. This is just a software version of the standard kinds of NLB found in cisco boxes.
-
Component Load Balancing. This is the most suitable. It's provided by Application Center and it allows you to deploy COM+ objects on a cluster of machines and have the calls distributed according to the load on those machines. You can control the threading and lifetime of the objects and view the status of the machines pretty easily using the Application Center MMC plugin (or SNMP, I believe). You'd have to wrap the computational part of your application into one or more COM objects. Once you've done that then you can create and call those objects in the cluster as if it were one machine - the clustering is transparent to the client application. I played around with AC a bit when it was in beta for a project that I was working on. We didn't go with it in the end because the design of our application ended up not requiring it (we just went with hardware load balancing), but it seemed like pretty cool technology - if you're into the whole COM thing. It has a really cool rolling deployment feature where you can redeploy your components (and/or IIS application if you have one) to your cluster incrementally while it's still running.
Here's some links to docs on MS's site:Introducing Windows 2000 Clustering Technologies
Application Center home page
Component Load Balancing
i happen to like front-row center on the upper level.
I looked at their site and they list at least 117 distinct patents. Does anyone know exactly which are in dispute?