It's easy to be critical of software engineering when all your doing is writing little op-ed pieces for Salon. What if (as a writer) you had to write an article that every single reader would understand without error? As a programmer that's what you're asking me to do.
Do you know why computer manufacturer X can't tell you why your modem didn't work? 'Cause it would probably take many times the cost of the machine you bought to test every single component, test all the drivers etc. It's just not feasible to do this. The complexity of the hardware alone in a modern PC is pretty mind boggleing. Your cpu has millions of transistors on it. Would you like me to test all those for you too Mr Salon? Are you nuts? When nasa had to upgrade the chips on the hubble telescope do you think the hubble got the latest whiz-bang pentium? No. It got a 486 (or something like that, I can't remember of the top of my head). Do you know why? Because NASA wanted to thoroughly test the chips before they put them up on the satelite. And after two years of rigorous testing they still didn't test every machine state. And you want me to "test" your system to tell you why your modem isn't working. What a joke.
Now onto the topic of software (your op-ed piece switches between the two fairly regulary, even though they are quite different. I will too.) Suppose you want to thoroughly test a program that allows a name (up to 20 chars) an address (20 chars also) and a phone number (10 digits). This program is so rediculessly simple testing it should be easy right? There are 26^20 combinations for the name field, 26^20 combinations of input for the address field, and 10^10 combinations for the phone number field. This gives a total number of possiblites as approximately 10^66.(I borrowed this example from Steve McConnell's Code Complete btw). Even if you miss one of these test cases you cannot authoritavely say that your code is bug free. So what if you take a sensible approach and test all they types of input, make sure your test cases cause all paths through a function to be executed, test all your boundary conditions, use some utilities to fill all the memory on your system with "data" so you can test for un-initialized variables etc, etc. What then? Move on and do the same to the operating system? The drivers? Ultimately the complexity of the whole system is so great that no-one has the resources to test everything.
I certainly agree that more rigourous quality assurance measures have to be put in place. I think open-source software/free software is certainly a step in the right direction, but ultimately the systems you're dealing with are insanely complex. What we need is new testing methodologies, not vague winging from someone from salon, just because they had a problem with their modem.
I have a degree in medical engineering, and I always thought that the EE people had it a bit tougher, but then I remember talking to some of them who thought some of the stuff we were doing was pretty hard also. The people I REALLY had respect for were the ones who did a double degree in EE and CS in 5 years. They had an average workload (in terms of contact hours) that was 1.5 times greater than all the other types of engineering. To those who did this, I salute you.
I agree, I've never been interested in downloading and testing the Netscape betas 'cause I've got mozilla. I don't hear people making any noise about the neoplanet "browser" ('cause this is a linux forum right?) - it's just IE imbedded with a heap of commercial crap plied on top. Same with NS6.
I seem to remember reading in an engineering text book that the standard unit of force in the imperial measurement system (feet, inches etc) is called a slug. That's funny.
I couldn't agree more. Greenspun has his head up his arse on this one. The XP people maintain that if you have to work MORE than a 40 hour week for 2 consecutive weeks then it is a sign that something is wrong with the management of the project. I love programming, but there are lots of things I love more, like spending time with my family. Also, on another not - isn't Greenspun the "one you publish something on the internet you should never remove it. Broken URLs will be punished by death" Guy. The URL for towzone (http://towzone.arsdigita.com) was broken. What gives?
Lots of microsoft DLLs are available for download from their web site as updates and patches (did anyone say DLL hell?), and I think most of the ole dlls (oleaut32.dll, olepro.dll etc) are redistributable if you have any of the microsoft developer tools. I wouldn't be surprised if office 2000 didn't come with a lot of the latest versions of these anyway. As a lot of other posters have mentioned M$ has to support a wider base of operating systems now, but really the windows version you install is pretty irrelevant once you start installing big packages (did somebody say bloatware) like visual studio, office 2K, IE5.x etc. Packages like this upgrade so many "system" dlls you're really looking at a whole new beast once a few of those have been installed.
I've been using the gimp for about 6 months on windows. It works fine. From what I've seen the development lags behind the Linux version a bit (I think I read that on the web site - I haven't needed to update it since I installed it), but I've never had a problem with it. If you're just bitching 'cause it uses the GTK+ widgets instead of the native windows ones then get over it. Just think of it as a very bland "skin" and you'll do fine.
So if photoshop is so easy to use why do I see so many photoshop books? (amazon returned 312 matches) Surely none of these would be necessary if photoshop was totally intuitive. I don't think any complex tool like the GIMP or Photoshop can every ditch the manual alltogether and rely totally on "intuition". Intuitive is what you're used to. If you're a mac person and that's what you're used to then the mac way of doing things is "intuitive". If you're a windows person then the windows way of doing things is intuitive (and possibly very frustrating to you). Of course if you're a professional artist who's used photoshop lots then it will be "intuitive". Cross-platform programs like photoshop and the gimp often can't leverage the experience users have on their particular platform, they have to aim for more common ground, or favour one platform over another.
I've been playing around with it for a couple of weeks and it works well. I really like the wxWindows bindings for python 2.0 (I also love the number of gui widget toolkits you can get - I was just looking at the fast, light toolkit the other day - it also has python bindings). The great thing about scripting languages and open source is that there's no need to ship binaries and source, the executable code and the source are one and the same thing. Whereas in the proprietary software world this would be a problem (you need to actually take steps to "protect" your code), in the open-source world this is an advantage.
Games are one of the driving forces that have pushed PC's from being monochrome text-based consoles to what we have today. So long as games companies keep demanding more state of the art machines for their games to run on, and so long as people keep wanting to play games (computer games industry was set to make more money than hollywood I believe this year) I think PC sales will still have some legs in 'em. Those folks up in Redmond don't mind jacking up the requirements for their OS to run either, so I can see that as a driving force behind continuing PC sales (for a couple of years more at least:)).
I wasn't aware of the MCHawking's prowess on the MIC, until today. Favourite Lyrics..
Dr. Dre can suck my dick,
that bitch got no PHD,
I lost count of mine,
I got stupid whack degrees.
Complex math it ain't no thing,
I'm mad dope crazy fly,
like Quantum formula,
I'll leave you asking why.
Certainly the work that our former coleague did may have caused some of our pain, but the language and/or IDE seems suspect in a few areas. It may be that we're still on the steep part of the learning curve (steep, sharp with broken glass and barbed wire in our path) but I've seen some pretty basic things just not work. I really hope Kylix is good, but I'm not buying into the hype (and that's all this is, there isn't really much content in the article) just yet. As to that crack about job security....well, that was a little below the belt. But I do know of a delphi programmer working in web development. He needed to do web pages that hooked up to a database back end. So what did he do? He re-implemented asp in delphi - poorly. He started about a year ago and is still going. His employer has started doing all their work in this beta quality asp workalike language. Hence my quip about job security.
I've always heard good things about delphi and naturally expected it to be a good programming language/development environment. Recently a member of our team left and the others in the group (non-delphi programmers) have started to have to do maintenance work on the code. The members of our team have a broad range of experience including VB, c++ and Java. Since our delphi programmer left management has resolved never to undertake another project in delphi ever again on pain of death due to the problems it has caused. Some members of our team are convinced that delphi programmers only continue to use it for the job security that it brings.
I laughed when I saw this article title
Highly Tuned NT Whips Barely Tuned Linux in MS-backed Test (reffering of course to the mindcraft fiasco from a while back). If only they had said this from the beginning.
I could see somthing like this working like HotDispatch or eLance, where someone poses a question like "I need info on somhting like X that existed before Y", if you have info on it then you get $25 bucks. I could turn out to be the cheapest prior art search around. Of course I think patents like the one we are discussing here are farcical and I'm glad people are comitted to seeing them overturned. I can't help but think that the only difference between the architecture that has been patented here and hundreds of thousands of mainframe apps is the GUI in place of the CLI. Surely since desktop applications were moving to GUI it is obvious that the interfaces to mainframe or server based applications would also move in this direction? Isn't there some provision in patent law saying that only things that are not obvious to someone skilled in the art can be patented?
It listens on port 1433 if I remember correctly. I'm not a h4X0r or anything like that, but I've developed some web sites for people using ASP (god, what I wouldn't have given for types, that's not asking too much is it?, anyway..) In the process I read some stuff on a few asp web sites and found info on an IIS exploit that (when not properly patched) allowed you to view asp code as plain text. That's no big deal, no one includes anything critical like passwords in their asp code do they? I was pretty amazed when I saw a fairly high-profile e-commerce web-site had their SQL Server passwords there in plain text. Not only that but they hadn't been changed from the default "sa". This was the first web-site I tried so I am sure it is not an isolated case. My point (I do have one) is that people who should know better, who spend millions on advertising, still can't set up their database correctly (or treat security as a product, not a process 'cause that would be too expensive). All the best software isn't worth a pinch of shit if you don't set it up correctly. I don't think anyone in the linux community would claim that linux is totally secure out of the box. The problem is how insecure SQL Server and a lot of other MS products are. Our sys-admin has a list of a couple of hundred things you have to do to make a fresh NT box somthing approximating "secure". I think the vulnerability in SQL is a real problem, and as you point out it has been a real problem for some time. There are hundreds, possibly thousands of web sites out there with a major security hole in them and you call it hysterical handwringing.
I heard a story from one of our sys-admins at work about a friend of his who submerged his CPU in a bath of non-conducting mineral oil which was chilled to sub-zero temperatures. I love the whole cryogenic computing thing.
I think the largest revolution in software engineering in the last 5 years or so has been design patterns. Components are great for GUI stuff, and pre-packaged classes are fine when they do just what you want. When somthing falls short you need to be able to develop your own stuff without re-inventing the wheel. I think this is where design patterns really shine, and I hope pattern libraries become more standard across languages.
Every time I hear people ripping on Mozilla I'm forced to speak up. I like Mozilla a lot. I've never had it crash on me (Unlike IE), it works quite well, and it provides a very important thing. Choice. How did Mozilla save my life? Well, I was writing an windows application with a web browser imbedded. Many of the clients we were writing this application for were adamant that they would NEVER install IE on their system (I think they're kidding themselves if they think they can keep it off their system running a windows box but anyway...) This was a big problem and looked as though it might de-rail the whole project, but then in stepped mozilla. Problem solved. It has never even crashed when running as an imbedded activeX control (you can even imbed mozilla in IE and let mozilla do all your html rendering in IE!). I think the future of mozilla is in hand-held devices, set top boxes etc, but I certainly don't think it's dead. Then again I guess it's easier for "journalists" to keep a job by tearing things down. A headline like "The mozilla developers are doing a great job, Thanks guys!" doesn't really make for great copy.
If I understand you correctly you need to expose the same bits of information in a number of different ways. Is it possible that XML might be able to solve these problems? You just write your information once and use differents XSL style sheets to expose the data in different ways (I know it's never that simple but..). If your data changes then any changes will automatically be reflected in your documents by re-generating them with XSL. That kind of changes the problem from being "how do I manage this information" to "what is an easy, flexible and quick way of writing XSL". I know XML is still a bit too close to the bleeding edge for commercial development, but I believe it holds a lot of promise.
I was looking at the list of features that microsoft intends to add to VB for the next release (VB7), and they are (allegedly) finally adding many of the features that are present in most 4GL's (inheritance, multi-threading). In the words of my brother (a java accolyte through and through) "I've never seen a language try and become java so quickly." Perhaps they will let you compile to java bytecode but how would this aid microsoft? They could make it only run on the Microsoft JVM, but why try and shackle VB programmers to Windows? They are already about as shackled to windows as anyone can be.
ASP is another snake-based language, but there is certainly nothing cool about it. CORBA is pretty close to COBRA, and is also pretty cool, so I guess we should mention it also.
I started doing some research about the I.T. jobs shortage, and found that on the surface there was almost universal agreement that yes there was a shortage and anyone who could put their pants on the right way could get a job in software. Then I went a bit deeper, and came across the work of one Dr.Norman Matloff. In his presentation that went before the U.S. House Judiciary Comittee he argues very convincinly that the only shortage is of young and relatively cheap programmers who are willing to work insane hours. Industry heavy-weights lobby government to allow them to "import" more cheap labour from India, eastern europe etc. It seems that (for once) when the laws of supply and demand are stacked in favour of the prospective employee the prospective employers out there felt they had to get together and change the rules.
Slashdot Mirror
I just re-read the article, all references in my above post to a "Mr Salon" shoud be ammended to "Miss Salon" or "Mrs Salon".
Do you know why computer manufacturer X can't tell you why your modem didn't work? 'Cause it would probably take many times the cost of the machine you bought to test every single component, test all the drivers etc. It's just not feasible to do this. The complexity of the hardware alone in a modern PC is pretty mind boggleing. Your cpu has millions of transistors on it. Would you like me to test all those for you too Mr Salon? Are you nuts? When nasa had to upgrade the chips on the hubble telescope do you think the hubble got the latest whiz-bang pentium? No. It got a 486 (or something like that, I can't remember of the top of my head). Do you know why? Because NASA wanted to thoroughly test the chips before they put them up on the satelite. And after two years of rigorous testing they still didn't test every machine state. And you want me to "test" your system to tell you why your modem isn't working. What a joke.
Now onto the topic of software (your op-ed piece switches between the two fairly regulary, even though they are quite different. I will too.) Suppose you want to thoroughly test a program that allows a name (up to 20 chars) an address (20 chars also) and a phone number (10 digits). This program is so rediculessly simple testing it should be easy right? There are 26^20 combinations for the name field, 26^20 combinations of input for the address field, and 10^10 combinations for the phone number field. This gives a total number of possiblites as approximately 10^66.(I borrowed this example from Steve McConnell's Code Complete btw). Even if you miss one of these test cases you cannot authoritavely say that your code is bug free. So what if you take a sensible approach and test all they types of input, make sure your test cases cause all paths through a function to be executed, test all your boundary conditions, use some utilities to fill all the memory on your system with "data" so you can test for un-initialized variables etc, etc. What then? Move on and do the same to the operating system? The drivers? Ultimately the complexity of the whole system is so great that no-one has the resources to test everything.
I certainly agree that more rigourous quality assurance measures have to be put in place. I think open-source software/free software is certainly a step in the right direction, but ultimately the systems you're dealing with are insanely complex. What we need is new testing methodologies, not vague winging from someone from salon, just because they had a problem with their modem.
I have a degree in medical engineering, and I always thought that the EE people had it a bit tougher, but then I remember talking to some of them who thought some of the stuff we were doing was pretty hard also. The people I REALLY had respect for were the ones who did a double degree in EE and CS in 5 years. They had an average workload (in terms of contact hours) that was 1.5 times greater than all the other types of engineering. To those who did this, I salute you.
I agree, I've never been interested in downloading and testing the Netscape betas 'cause I've got mozilla. I don't hear people making any noise about the neoplanet "browser" ('cause this is a linux forum right?) - it's just IE imbedded with a heap of commercial crap plied on top. Same with NS6.
I seem to remember reading in an engineering text book that the standard unit of force in the imperial measurement system (feet, inches etc) is called a slug. That's funny.
I couldn't agree more. Greenspun has his head up his arse on this one. The XP people maintain that if you have to work MORE than a 40 hour week for 2 consecutive weeks then it is a sign that something is wrong with the management of the project. I love programming, but there are lots of things I love more, like spending time with my family. Also, on another not - isn't Greenspun the "one you publish something on the internet you should never remove it. Broken URLs will be punished by death" Guy. The URL for towzone (http://towzone.arsdigita.com) was broken. What gives?
Lots of microsoft DLLs are available for download from their web site as updates and patches (did anyone say DLL hell?), and I think most of the ole dlls (oleaut32.dll, olepro.dll etc) are redistributable if you have any of the microsoft developer tools. I wouldn't be surprised if office 2000 didn't come with a lot of the latest versions of these anyway. As a lot of other posters have mentioned M$ has to support a wider base of operating systems now, but really the windows version you install is pretty irrelevant once you start installing big packages (did somebody say bloatware) like visual studio, office 2K, IE5.x etc. Packages like this upgrade so many "system" dlls you're really looking at a whole new beast once a few of those have been installed.
I've been using the gimp for about 6 months on windows. It works fine. From what I've seen the development lags behind the Linux version a bit (I think I read that on the web site - I haven't needed to update it since I installed it), but I've never had a problem with it. If you're just bitching 'cause it uses the GTK+ widgets instead of the native windows ones then get over it. Just think of it as a very bland "skin" and you'll do fine.
So if photoshop is so easy to use why do I see so many photoshop books? (amazon returned 312 matches) Surely none of these would be necessary if photoshop was totally intuitive. I don't think any complex tool like the GIMP or Photoshop can every ditch the manual alltogether and rely totally on "intuition". Intuitive is what you're used to. If you're a mac person and that's what you're used to then the mac way of doing things is "intuitive". If you're a windows person then the windows way of doing things is intuitive (and possibly very frustrating to you). Of course if you're a professional artist who's used photoshop lots then it will be "intuitive". Cross-platform programs like photoshop and the gimp often can't leverage the experience users have on their particular platform, they have to aim for more common ground, or favour one platform over another.
I've been playing around with it for a couple of weeks and it works well. I really like the wxWindows bindings for python 2.0 (I also love the number of gui widget toolkits you can get - I was just looking at the fast, light toolkit the other day - it also has python bindings). The great thing about scripting languages and open source is that there's no need to ship binaries and source, the executable code and the source are one and the same thing. Whereas in the proprietary software world this would be a problem (you need to actually take steps to "protect" your code), in the open-source world this is an advantage.
Games are one of the driving forces that have pushed PC's from being monochrome text-based consoles to what we have today. So long as games companies keep demanding more state of the art machines for their games to run on, and so long as people keep wanting to play games (computer games industry was set to make more money than hollywood I believe this year) I think PC sales will still have some legs in 'em. Those folks up in Redmond don't mind jacking up the requirements for their OS to run either, so I can see that as a driving force behind continuing PC sales (for a couple of years more at least :)).
Dr. Dre can suck my dick,
that bitch got no PHD,
I lost count of mine,
I got stupid whack degrees.
Complex math it ain't no thing,
I'm mad dope crazy fly,
like Quantum formula,
I'll leave you asking why.
I noticed win2K wasn't on the list. Hmmm....
Certainly the work that our former coleague did may have caused some of our pain, but the language and/or IDE seems suspect in a few areas. It may be that we're still on the steep part of the learning curve (steep, sharp with broken glass and barbed wire in our path) but I've seen some pretty basic things just not work. I really hope Kylix is good, but I'm not buying into the hype (and that's all this is, there isn't really much content in the article) just yet. As to that crack about job security....well, that was a little below the belt. But I do know of a delphi programmer working in web development. He needed to do web pages that hooked up to a database back end. So what did he do? He re-implemented asp in delphi - poorly. He started about a year ago and is still going. His employer has started doing all their work in this beta quality asp workalike language. Hence my quip about job security.
I've always heard good things about delphi and naturally expected it to be a good programming language/development environment. Recently a member of our team left and the others in the group (non-delphi programmers) have started to have to do maintenance work on the code. The members of our team have a broad range of experience including VB, c++ and Java. Since our delphi programmer left management has resolved never to undertake another project in delphi ever again on pain of death due to the problems it has caused. Some members of our team are convinced that delphi programmers only continue to use it for the job security that it brings.
I laughed when I saw this article title Highly Tuned NT Whips Barely Tuned Linux in MS-backed Test (reffering of course to the mindcraft fiasco from a while back). If only they had said this from the beginning.
I could see somthing like this working like HotDispatch or eLance, where someone poses a question like "I need info on somhting like X that existed before Y", if you have info on it then you get $25 bucks. I could turn out to be the cheapest prior art search around. Of course I think patents like the one we are discussing here are farcical and I'm glad people are comitted to seeing them overturned. I can't help but think that the only difference between the architecture that has been patented here and hundreds of thousands of mainframe apps is the GUI in place of the CLI. Surely since desktop applications were moving to GUI it is obvious that the interfaces to mainframe or server based applications would also move in this direction? Isn't there some provision in patent law saying that only things that are not obvious to someone skilled in the art can be patented?
It listens on port 1433 if I remember correctly. I'm not a h4X0r or anything like that, but I've developed some web sites for people using ASP (god, what I wouldn't have given for types, that's not asking too much is it?, anyway..) In the process I read some stuff on a few asp web sites and found info on an IIS exploit that (when not properly patched) allowed you to view asp code as plain text. That's no big deal, no one includes anything critical like passwords in their asp code do they? I was pretty amazed when I saw a fairly high-profile e-commerce web-site had their SQL Server passwords there in plain text. Not only that but they hadn't been changed from the default "sa". This was the first web-site I tried so I am sure it is not an isolated case. My point (I do have one) is that people who should know better, who spend millions on advertising, still can't set up their database correctly (or treat security as a product, not a process 'cause that would be too expensive). All the best software isn't worth a pinch of shit if you don't set it up correctly. I don't think anyone in the linux community would claim that linux is totally secure out of the box. The problem is how insecure SQL Server and a lot of other MS products are. Our sys-admin has a list of a couple of hundred things you have to do to make a fresh NT box somthing approximating "secure". I think the vulnerability in SQL is a real problem, and as you point out it has been a real problem for some time. There are hundreds, possibly thousands of web sites out there with a major security hole in them and you call it hysterical handwringing.
I heard a story from one of our sys-admins at work about a friend of his who submerged his CPU in a bath of non-conducting mineral oil which was chilled to sub-zero temperatures. I love the whole cryogenic computing thing.
I think the largest revolution in software engineering in the last 5 years or so has been design patterns. Components are great for GUI stuff, and pre-packaged classes are fine when they do just what you want. When somthing falls short you need to be able to develop your own stuff without re-inventing the wheel. I think this is where design patterns really shine, and I hope pattern libraries become more standard across languages.
Every time I hear people ripping on Mozilla I'm forced to speak up. I like Mozilla a lot. I've never had it crash on me (Unlike IE), it works quite well, and it provides a very important thing. Choice. How did Mozilla save my life? Well, I was writing an windows application with a web browser imbedded. Many of the clients we were writing this application for were adamant that they would NEVER install IE on their system (I think they're kidding themselves if they think they can keep it off their system running a windows box but anyway...) This was a big problem and looked as though it might de-rail the whole project, but then in stepped mozilla. Problem solved. It has never even crashed when running as an imbedded activeX control (you can even imbed mozilla in IE and let mozilla do all your html rendering in IE!). I think the future of mozilla is in hand-held devices, set top boxes etc, but I certainly don't think it's dead. Then again I guess it's easier for "journalists" to keep a job by tearing things down. A headline like "The mozilla developers are doing a great job, Thanks guys!" doesn't really make for great copy.
If I understand you correctly you need to expose the same bits of information in a number of different ways. Is it possible that XML might be able to solve these problems? You just write your information once and use differents XSL style sheets to expose the data in different ways (I know it's never that simple but..). If your data changes then any changes will automatically be reflected in your documents by re-generating them with XSL. That kind of changes the problem from being "how do I manage this information" to "what is an easy, flexible and quick way of writing XSL". I know XML is still a bit too close to the bleeding edge for commercial development, but I believe it holds a lot of promise.
I was looking at the list of features that microsoft intends to add to VB for the next release (VB7), and they are (allegedly) finally adding many of the features that are present in most 4GL's (inheritance, multi-threading). In the words of my brother (a java accolyte through and through) "I've never seen a language try and become java so quickly." Perhaps they will let you compile to java bytecode but how would this aid microsoft? They could make it only run on the Microsoft JVM, but why try and shackle VB programmers to Windows? They are already about as shackled to windows as anyone can be.
ASP is another snake-based language, but there is certainly nothing cool about it. CORBA is pretty close to COBRA, and is also pretty cool, so I guess we should mention it also.
I started doing some research about the I.T. jobs shortage, and found that on the surface there was almost universal agreement that yes there was a shortage and anyone who could put their pants on the right way could get a job in software. Then I went a bit deeper, and came across the work of one Dr.Norman Matloff. In his presentation that went before the U.S. House Judiciary Comittee he argues very convincinly that the only shortage is of young and relatively cheap programmers who are willing to work insane hours. Industry heavy-weights lobby government to allow them to "import" more cheap labour from India, eastern europe etc. It seems that (for once) when the laws of supply and demand are stacked in favour of the prospective employee the prospective employers out there felt they had to get together and change the rules.