SQLServerCentral / Editorials / SQLServerCentral.com / Building Better Software / Latest Posts

RE: Building Better Software

Steve Jones - SSC Editor — Sun, 27 Jan 2013 10:56:35 GMT

Good points, Scott, and I wasn't attempting to equate software to building, though they both do poor jobs, so maybe that isn't so bad. Most buildings aren't build completely square or plumb, or on time, or even without "bugs" that are found later.But I agree with you, it's a tenuous analogy. The idea of medicine does make more sense, though in that case doctors make mistakes constantly, despite knowing better.I think we can engineer things to a point, but much of what we build becomes beyond the ability of one person to understand and comprehend the problems. Testing is poorly done, despite the fact that it's gotten much better, and much more comprehensive, it doesn't even come close to the real world.

RE: Building Better Software

Lynn Pettis — Fri, 25 Jan 2013 22:33:07 GMT

And I have to disagree with you. Software can be engineered, just like any other project. Whether or not it actually is happens to be another people problem, not a software problem.

RE: Building Better Software

Scott Anderson #2 — Fri, 25 Jan 2013 20:40:08 GMT

Funny side point: I just tried to post this reply (just above as I'm editing the original out now), but the send timed out. I refresh the page and all my text is gone! Arhhh I'm just speaking to the wind! Should I tell someone that there is a bug? Who, where, why, how do they reproduce it? If I didn't take my own precaution by writing it first in another application, proof read and then copy and paste it in here again, I would have lost it all. No, there is no better thing the website developers could have done, I have to take responsibility and treat software as what it is; just a very fallible tool.

RE: Building Better Software

Scott Anderson #2 — Fri, 25 Jan 2013 20:34:31 GMT

Arhh Steve not you too; thinking software *creation* is analogous to architectural building. That was only the selling point to get people to buy these systems in the first place many years ago. Nothing could be further from the truth, but you don’t dare tell the financial stake holders that. Especially such facts as 70% of software projects fail. How many building projects does that happen to? Got to associate with the right mobs to get the risk taken and get paid.Writing software is *nothing* like building houses. Funny enough the closest profession is… (believe it or not, it will take a bit but it will click)… Doctors. Huh? Really? No way! Now you’re just making things up!?!Doctors have to deal with the following:1) ever changing organisms (virus, bacteria, cell mutations, etc.)2) can only ever try what they *think* has worked before3) have to deal with the humans who cause the problems to themselves and others4) no two patients are exactly alike even though they are all humans5) and on and on that similarities go but who has a couple of hours to spare?Now if you view software development from that point of view, then you have a more realistic understand of what can and cannot be done to build better software.

RE: Building Better Software

Scott Anderson #2 — Fri, 25 Jan 2013 19:59:19 GMT

Everyone having thick skin makes getting to the interesting points quicker. Good.Ok Gary, I didn’t want to tangent off to your common view as that rabbit hole goes real deep, but I think the human factor point has been made.Software is repeatable: Yes, given a functional programming language, with a scope that is immutable and cannot differ, hardware that cannot change, the absolutely yes; otherwise no chance in bucklies it’s repeatable. Read on...Software is testable: Really? Fact: Only a surprisingly small part of software is testable. Have you seen Windows, Office, SQL Server, Linux, etc? Are you certain that with the amazing resources they have, that they don’t test their software? Automated, integrated, user acceptance and a myriad of other techniques, and yet somehow they release a surprising number of patches and fixes so often. Read on...Why do you think software developers talk in terms of “Should”, “Depends”, “May”, “Probably”, “Hope”? Because there are no guarantees. To justify the pay they need titles like Software Engineer, not Software Artist, which is really what they are doing. The only “engineering” is when they use years proven 3rd party libraries. Eg. Do you know how often code fails when getting data from a database? Rarely ever, unless some user does something stupid. It’s everything else that happens around it which is “done for the very first time” to communicate with the user that has the problems. Every application is different and there are very few repeatable parts, otherwise you would only ever have a couple of programs. This is why there are so few database access libraries, to have some stability somewhere.Should occur before release: Sorry but huh? Of course it does! Every heard the saying “Works on my machine” or “I can’t reproduce your problem”? Every computer has a different hardware setup, has different combination of applications and drivers installed. There is no set know environment. It’s all different! How can you find problems in software that your machine doesn’t have?It would be great it software was as easy as drag/drop and write a bunch of lines. Imaging how fast applications and systems could be created!?! But funny enough, software development isn’t fast and it takes time, so how could that be? Arhh those hobbyist, they are the ones that make new versions of Windows, Office and SQL Server take years to come around. Let’s get them!No, the realities are that like I said, it’s nigh to impossible to better engineer software with the current tools and more importantly; it is in no way easy to develop software with any complexity. Ie. Interacting with the user.BUT! We can do better in other ways…So Gary, please don’t think I’m picking on you, it’s just you’ve articulated a common mis-conception and fallacy; so you’ve been made an example of. ;-)

RE: Building Better Software

Steve Jones - SSC Editor — Fri, 25 Jan 2013 10:28:40 GMT

RE: Building Better Software

Gary Varga — Fri, 25 Jan 2013 04:11:43 GMT

Also, this is a discussion. Feathers are there to be ruffled in open debate ;-)

RE: Building Better Software

Gary Varga — Fri, 25 Jan 2013 04:10:41 GMT

[quote][b]Scott Anderson-466019 (1/24/2013)[/b][hr]Sorry if I ruffled a few feathers, but I always question peoples beliefs on why they think software is any more special than any other tool they use. That being said, I also understand how they came to that belief. There is nothing like software. You can’t touch, hold, manipulate without more software or via the developer and when it works, it works very well, so it’s mystical in nature.But (there is always a but) as Old Hand says software of any reasonable complexity is very hard to get working, let alone working to what everyone thinks is right. Don’t think so? Have a try with something simple like a recommendation engine. This is why I think Steve’s line of “Developing software is easy…” irks me, even though he correctly tries to explain it away. “Hello World” is easy; everything after isn’t easy when you have human users.As Lyn says “trap for mistakes”, well that’s easier said than done. Look at the legal systems in their complexity and loop-hole flaws by trying to trap for mistakes.As Gary unintentionally agrees, cars and microwaves fail (more often because of the operator) but we don’t say “the manufacture should have tested for my problem case better!”So building better software is extremely hard and nigh to impossible. Just saying more testing won't make that happen. The varied skill of the users, the varied conditions it's used in, the pressure of getting it to market, the change requirements expected; it's a wonder it even works at all!The bottom line is if you look at why you are frustrated by feeling powerless with buggy software is because of the mystical nature in that with other tools you can somewhat fix the problem yourself. Car gets a flat; you change the tire; microware stops working you buy another one, light bulb; replace, but bug in software… can’t easily replace, fix, *arrhggg* call the mighty developers and hope for the best. [/quote]Software is not "any more special" but it is a repeatable, testable artefact that allows for refinement. The case in point in this editorial is that refinement, through software engineering techniques say, should occur BEFORE release. Ironically, it is the human decision not to apply the efforts in what I believe is the appropriate time that causes the problem. It is the repeatability that I believe invalidates your nullification of your dismissal of Lynn's viewpoint; legal systems are not digital and are totally based on human judgement thus making it not repeatable.I think you are being a little disingenuous with some of your comments, particularly by quoting half of Steve's sentence which completely alters the intent of the message (even with your qualification). The whole point is that IT IS EASY to develop software. Drag and drop your UI then write a bunch of lines which now can be butchered versions of what has been copied from the Internet. Surely you have been collared by hobbyist coders. The message was that it isn't easy to do a good job....and no I don't agree. Intentionally. I do tell manufacturers that their products should work as advertised. This isn't an edge case as far as the user is concerned. The days referred to in the article are the same as any other day. It was an edge case of the component used. Something a professional should know to cater for.Also your bottom line is not comparing like for like. A flat tyre is expected with a solution designed in. It would be fairer to compare a new virus discovered and an antivirus software company supplying an updated virus definitions file.NOTE TO HOBBYIST CODERS: I have nothing against you as it is a hobby. I just think that there is a separate level of requirement for professionals. Please see my (lack of) sporting abilities that I regularly demonstrate for evidence that I believe that anyone can do these things but you have to be really good to be professional. I also apply plasters to my kids but I will not perform surgery on them ;-)

RE: Building Better Software

Scott Anderson #2 — Thu, 24 Jan 2013 17:06:59 GMT

Sorry if I ruffled a few feathers, but I always question peoples beliefs on why they think software is any more special than any other tool they use. That being said, I also understand how they came to that belief. There is nothing like software. You can’t touch, hold, manipulate without more software or via the developer and when it works, it works very well, so it’s mystical in nature.But (there is always a but) as Old Hand says software of any reasonable complexity is very hard to get working, let alone working to what everyone thinks is right. Don’t think so? Have a try with something simple like a recommendation engine. This is why I think Steve’s line of “Developing software is easy…” irks me, even though he correctly tries to explain it away. “Hello World” is easy; everything after isn’t easy when you have human users.As Lyn says “trap for mistakes”, well that’s easier said than done. Look at the legal systems in their complexity and loop-hole flaws by trying to trap for mistakes.As Gary unintentionally agrees, cars and microwaves fail (more often because of the operator) but we don’t say “the manufacture should have tested for my problem case better!”So building better software is extremely hard and nigh to impossible. Just saying more testing won't make that happen. The varied skill of the users, the varied conditions it's used in, the pressure of getting it to market, the change requirements expected; it's a wonder it even works at all!The bottom line is if you look at why you are frustrated by feeling powerless with buggy software is because of the mystical nature in that with other tools you can somewhat fix the problem yourself. Car gets a flat; you change the tire; microware stops working you buy another one, light bulb; replace, but bug in software… can’t easily replace, fix, *arrhggg* call the mighty developers and hope for the best.

RE: Building Better Software

TravisDBA — Thu, 24 Jan 2013 10:33:26 GMT

Scenario One: Plane crashes due to human error. Scenario Two: Plane crashes due to computer software error. Personally, if I were on that plane, it wouldn't really concern me what caused it, or the difference between the two for that matter. Bottom line is I'd probably be dead either way. I think we all are comparing oranges to tangerines here. Just my take...:-D

RE: Building Better Software

Steve Jones - SSC Editor — Thu, 24 Jan 2013 10:04:56 GMT

Scott,There's something to your argument, but I'm not sure how to exactly examine it. As Gary mentioned, poorly written software, or software with issues doesn't necessarily relate when talking about the problems with our use of computers. We do have problems with computers, and software can help reduce some of those, but education, practice, retaining people, will also help. That's a separate problem.

RE: Building Better Software

Miles Neale — Thu, 24 Jan 2013 09:46:32 GMT

[quote][b]Gary Varga (1/24/2013)[/b][hr]Sorry Scott but I agree with Steve and Lynn.[/quote]I guess that I agree with all three of you. There are two problems being discussed and both are really problems developers face. When there are errors, bugs, or defects in the software and it does not execute as expected that is a problem. When people use software it can be really messy and they can introduce errors. The real issue in the development of software is the development of a product that will do what is expected when the correct data and procedure are followed. And at the same time when the human element is introduced to the process it is inherent in the development of the software to "guide" the user into doing it right. Historically we have people who have called this business Software Engineering. That is not just the applied science of Software Development but it is also the engineering of a process built into the software to motivate or entice the user to do it right. In the development of software we often leave the correct path so cryptic and unclear that the user is at a disadvantage and is lucky if they get it right. We can eliminate some human error, not all and we have to remember but we can eliminate some human error with clearly defined pathways through our software and automating the solutions such that the process tells the user that they are about to really blow it or that they are doing it right.Problem is that if we seek to develop a solution that is without defect and the process and procedure is perfect we spend far above the normal allowable resources and get into financial problems in our projects. If I were to state what I think the problem is I would say that the compromise between what we are willing to pay for and what we need to build it correctly, far too often causes the usability and performance of the software to suffer.M.

RE: Building Better Software

Gary Varga — Thu, 24 Jan 2013 04:08:26 GMT

Sorry Scott but I agree with Steve and Lynn.In the context of a website such as this and an article titled "Building Better Software" we are not comparing the various causes and their produced effects we are talking about engineering better software. Your comparison is valid but, in this context, irrelevant. People are infuriated on a daily basis by poorly written software and, justifiably in my opinion, expect better.They feel the same about other products; cars should rarely break down and microwaves should always cook etc.There should be, and have been, discussions and articles on the human causes failures. This did not read like one to me (enjoyed it though, thanks Steve).

RE: Building Better Software

Gary Varga — Thu, 24 Jan 2013 03:58:53 GMT

[quote][b]sruthi.kumar (1/23/2013)[/b][hr]Sometimes I wonder why larger organizations would not test their products before release. Often, I get answer like they want their customers to find bugs. <snip/>[/quote]I have heard people in organisations that they will put up with customers finding defects as it makes more commercial sense to release earlier and take the flak rather than wait but as desired strategy? Never!!!BTW, I do not like the term "bug" as this terminology suggests that it is caused by something extraneous or external to the software. I know that most people who use the term don't feel this way but sometimes that is the impression left with users (and other developers). I prefer the term defect. That is what it is. Someone has created a defect in the software either through action or inaction. I, my colleagues, my peers and hobbyists (unintentionally) introduce defects and we must all accept that. It is human action that causes it not some unidentified invertebrate.

RE: Building Better Software

Lynn Pettis — Wed, 23 Jan 2013 21:31:41 GMT

[quote][b]Scott Anderson-466019 (1/23/2013)[/b][hr]Ok, not a convincing enough argument? How about the Internet. The biggest usage of software by humans and the major outages have been caused by what or whom? Human operator error.But back to your point: Software mistakes affect many people. Why is that? How have been become so dependent on software when it has so many problems and bugs? Because software bugs are a stereotype.Really, the amount of problems compared to working solutions is incomparable, but we pick on and declare the end of days when something doesn’t go as *we* expected. “Expected”, how many times has a new staff member or college not done things as we expected, but to write an article about that? That’s common knowledge and expected, so who cares!?!I agree that comparing human mistakes to software mistakes isn’t fair, as humans make so many more.[/quote]I'm lost with your argument. Sorry, just not making too much sense to me. Building better software really doesn't have much to do with some of the human errors you mentioned. I really think it is about writing software that does what is expected when properly used. And as we know, there are users that don't use software the way it was intended many times. If you can trap for those "mistakes" and handle them, then you are ahead of the curve in many respects.

RE: Building Better Software

Scott Anderson #2 — Wed, 23 Jan 2013 17:55:23 GMT

Ok, not a convincing enough argument? How about the Internet. The biggest usage of software by humans and the major outages have been caused by what or whom? Human operator error.But back to your point: Software mistakes affect many people. Why is that? How have been become so dependent on software when it has so many problems and bugs? Because software bugs are a stereotype.Really, the amount of problems compared to working solutions is incomparable, but we pick on and declare the end of days when something doesn’t go as *we* expected. “Expected”, how many times has a new staff member or college not done things as we expected, but to write an article about that? That’s common knowledge and expected, so who cares!?!I agree that comparing human mistakes to software mistakes isn’t fair, as humans make so many more.

RE: Building Better Software

Michael Valentine Jones — Wed, 23 Jan 2013 17:22:20 GMT

The referenced article also has plenty of mistakes. :laugh:Ask Ars: Why will Apple's Do Not Disturb bug fix itself next week?[url]http://arstechnica.com/apple/2013/01/ask-ars-why-will-apples-do-not-disturb-bug-fix-itself-next-week/[/url]For example:"The first day of 2013 started on a Tuesday, whereas (as noted by TUAW) the ISO standard expects the first week of the year to start on "the Monday that contains the first Thursday in January." In this case, that would be [u]January 7, 2013[/u]."The first ISO week for 2013 starts on December 31, 2012, not January 7, 2013. The first ISO week of the year can start no later that January 4 of any year, and will be in the range of 29 December of the prior calendar year to 4 January of the current calendar year.ISO week date[url]http://en.wikipedia.org/wiki/ISO_week_date[/url]

RE: Building Better Software

Steve Jones - SSC Editor — Wed, 23 Jan 2013 16:45:11 GMT

I'm not sure that comparing human mistakes to software mistakes is fair. Software mistakes can easily affect a large group of people and impact lots of business. Forgetting backups, while tragic, happens lots without there being a problem.I'm not sure which one is worse. Certainly there is something to the idea that humans reduce efficiency more than poorly written software, but that might be hard to quantify. Maybe I need to think about that a bit.

RE: Building Better Software

Scott Anderson #2 — Wed, 23 Jan 2013 16:37:43 GMT

Building Better Software or Bugs in Software... Unfortunately Steve you have put software on a far too high pedestal. Software, even cowboy hacker buggy software is more reliable than the human operators and users. If you compare software error/bugs in software to human error/bugs, you might have written a different article. Financial and productivity loss from human problems makes software problems an after thought. What you really should be writing about is how we falsely reason that it's so much easier to fix the human problem than the software problem, when really it's not.Ever had a human run a Delete with no where, not tested their backups, damaged the backups, turned the wrong server off, changed the sa password then gone one holidays, rogue admin, etc. That's just the easy ones. How about made wrong security choices, wrong storage location or space allocation, poor performing SQL, bad indexes, office politics, etc.Complaining about software bugs like it's a real source of drama is like complaining about the weather; it's just vanity.

RE: Building Better Software

sruthi.kumar — Wed, 23 Jan 2013 14:42:19 GMT

Sometimes I wonder why larger organizations would not test their products before release. Often, I get answer like they want their customers to find bugs. By the time customers start reporting bugs, don't the organizations think their reputation was already got damaged? Most of the times, products are not even beta versions. I think product needs to be well tested for at least obvious defects before even releasing to the market.

RE: Building Better Software

Miles Neale — Wed, 23 Jan 2013 11:54:46 GMT

It takes little understanding and talent to write bad code that will not work. That is easy. And for the person who has written a few hundred thousand lines of code, it might appear easy to throw something together that may or may not work. But the premise in the editorial is correct, it is not easy to get it right. Compare it to throwing together a few burgers or an omelet to a five course formal dinner with the Queen. One is very easy, the other requires attention to detail that most of us are not use to.Good experienced developers who can get the job done right and make it last are expensive. But the industry has to face the fact that those dime-a-dozen developers appear to be cheap to get a product out, but in the long run a far more expensive then the experienced professional. M.

RE: Building Better Software

Gary Varga — Wed, 23 Jan 2013 08:06:29 GMT

On top of all the above is the lack of demanding and/or allowing of following decent processes to ensure high quality software engineering occurs. We all know the "just ship it" attitude remains alive and well and still comes from people who know better but feel that the commercial pressures are too high.

RE: Building Better Software

roger.plowman — Wed, 23 Jan 2013 06:58:15 GMT

I've been a developer for over 30 years. No, it is [i]not[/i] easy! :-DThe problem is the size of the problem domain and the sheer complexity of it. Take the ISO date example. Who has time to be an expert on every single aspect of ISO dates? Being blissfully ignorant on that subject :hehe: my first question as a developer is "why doesn't the ISO date system operate as expected?" Is there a bug in it? If so, why? What's the problem and why would it bite on Jan 1st to 7th of 2013? And, being this is ISO approved, why would there be a bug at all???? If you can't trust the ISO who can you trust?Take my current developement project. I'm implementing what could best be described as a "whole business automation system", using VB.NET for the front end and T-SQL for the backend. That's two entire, and very complex, languages I have to know. Further, I've only known these languages for about a year now, while keeping up with all the other jobs I have to do as our company's one man IT shop.The point being I already know I'm [i]not[/i] fluent in either VB.NET or T-SQL and probably won't be for about 3 more years. By "fluent" I mean "know the languages inside out and backwards, able to wring every bit of performance and know all the gotchas to avoid". In the meantime I've got to create a program with a predecesser that consisted of 100,000 lines of code, written in a system I've been fluent in for 10 years. Oh, and did I mention my users won't take the time to beta test the new system? Because they don't have "time to play". Oh, and that my tool budget is severely constrained?And let's not even talk about all the (very complex) tools involved in production, from the ERD tool, to all the Red Gate tools, SSMS, Visual Studio (and add-ons), the profiler, the unit test system...Easy my Aunt Sally! :-PSo what do you do in that situation? You keep it simple. You avoid all the nifty tricks that would probably be faster or need less code, but have unforseen "gotchas" and inexplicable interactions with other features. Since you [i]can't[/i] know everything, you concentrate on a small subset you know won't break. Then you refactor as you can over the lifetime of the system. (The last system lasted 13 years!)And that, my childen, is how Equestria was made! (laughing)

Building Better Software

Steve Jones - SSC Editor — Tue, 22 Jan 2013 21:13:35 GMT

Comments posted to this topic are about the item [B]Building Better Software[/B]