Related
Somewhere I saw some personal notes from someone who was on the ANSI committee. I thought it was Kent Pitman, but a search of his site doesn't turn up anything. Neither does Google.
I'm interested in the background of the decision not to integrate the condition system with CLOS. CLtL2 speaks of it as a fait accompli, and I'm curious as to why it didn't happen.
The condition system was not integrated with CLOS because there were implementations with existing condition systems which were not CLOS-based (they were, in at least one case, flavors-based), because CLOS did not exist at all until pretty late in the standardisation process. Since a condition system has really deep roots in any implementation, requiring those implementations to rip out a great part of their guts in and replace them with some CLOS-based guts would have placed them – the very implementations which had gone out of their way to make sophisticated condition handling possible in the first place – at a huge disadvantage. Doing that would have been both stupid and would have derailed the standardisation process, since the representatives of those implementations would have been considerably antagonised by a decision like that. So the right decision was made.
It was also unclear at the time that CLOS could be made really performant on stock hardware (perhaps this is still unclear, but stock hardware is now so fast and we all happily live with implementations of other languages which are hugely slower than a good CLOS implementation can be so the problem no longer matters). CL was also considered really big (hard to remember when my fully-fledged hairy CL IDE containing the entire hyperspect and all its own documentation is 2/3 the size of my web browser), so people thought about subset implementations which might not contain CLOS but really needed to contain the condition system.
In particular it is worth looking at the CLHS issue (not part of the spec) CLOS-CONDITIONS-AGAIN, from which comes the following text:
The condition system should not be too tightly integrated into CLOS, for two reasons: Some implementations already have a native condition system that is not based on CLOS, and it should be possible to integrate the native conditions and the ANSI CL conditions. Some people would like to define an ANSI Common Lisp subset that does not contain CLOS but does contain conditions.
The problem areas are the use of DEFCLASS, MAKE-INSTANCE, and DEFMETHOD to define and create conditions, rather than using more abstract macros that conceal the implementation of conditions in terms of CLOS, and exposure of the implementation of condition slots as CLOS slots. If user code was written in a more abstract way, it could run in a subset language that did not contain CLOS.
This is not normative text but you can see what people were thinking.
I am trying to understand all the low-level stuff Compilers / Interpreters / the Kernel do for you (because I'm yet another person who thinks they could design a language that's better than most others)
One of the many things that sparked my curiosity is Async-Await.
I've checked the under-the-hood implementation for a couple languages, including C# (the compiler generates the state machine from sugar code) and Rust (where the state machine has to be implemented manually from the Future trait), and they all implement Async-Await using state machines.
I've not found anything useful by googling ("async copy stack frame" and variations) or in the "Similar questions" section.
To me, this method seems rather complicated and overhead-heavy;
Could you not implement Async-Await by simply memcopying the stack frames of async calls to/from heap?
I'm aware that it is architecturally impossible for some languages (I thank the CLR can't do it, so C# can't either).
Am I missing something that makes this logically impossible? I would expect less complicated code and a performance boost from doing it that way, am I mistaken? I suppose when you have a deep stack hierarchy after a async call (eg. a recursive async function) the amount of data you would have to memcopy is rather large, but there are probably ways to work around that.
If this is possible, then why isn't it done anywhere?
Yes, an alternative to converting code into state machines is copying stacks around. This is the way that the go language does it now, and the way that Java will do it when Project Loom is released.
It's not an easy thing to do for real-world languages.
It doesn't work for C and C++, for example, because those languages let you make pointers to things on the stack. Those pointers can be used by other threads, so you can't move the stack away, and even if you could, you would have to copy it back into exactly the same place.
For the same reason, it doesn't work when your program calls out to the OS or native code and gets called back in the same thread, because there's a portion of the stack you don't control. In Java, project Loom's 'virtual threads' will not release the thread as long as there's native code on the stack.
Even in situations where you can move the stack, it requires dedicated support in the runtime environment. The stack can't just be copied into a byte array. It has to be copied off in a representation that allows the garbage collector to recognize all the pointers in it. If C# were to adopt this technique, for example, it would require significant extensions to the common language runtime, whereas implementing state machines can be accomplished entirely within the C# compiler.
I would first like to begin by saying that this answer is only meant to serve as a starting point to go in the actual direction of your exploration. This includes various pointers and building up on the work of various other authors
I've checked the under-the-hood implementation for a couple languages, including C# (the compiler generates the state machine from sugar code) and Rust (where the state machine has to be implemented manually from the Future trait), and they all implement Async-Await using state machines
You understood correctly that the Async/Await implementation for C# and Rust use state machines. Let us understand now as to why are those implementations chosen.
To put the general structure of stack frames in very simple terms, whatever we put inside a stack frame are temporary allocations which are not going to outlive the method which resulted in the addition of that stack frame (including, but not limited to local variables). It also contains the information of the continuation, ie. the address of the code that needs to be executed next (in other words, the control has to return to), within the context of the recently called method. If this is a case of synchronous execution, the methods are executed one after the other. In other words, the caller method is suspended until the called method finishes execution. This, from a stack perspective fits in intuitively. If we are done with the execution of a called method, the control is returned to the caller and the stack frame can be popped off. It is also cheap and efficient from a perspective of the hardware that is running this code as well (hardware is optimised for programming with stacks).
In the case of asynchronous code, the continuation of a method might have to trigger several other methods that might get called from within the continuation of callers. Take a look at this answer, where Eric Lippert outlines the entirety of how the stack works for an asynchronous flow. The problem with asynchronous flow is that, the method calls do not exactly form a stack and trying to handle them like pure stacks may get extremely complicated. As Eric says in the answer, that is why C# uses graph of heap-allocated tasks and delegates that represents a workflow.
However, if you consider languages like Go, the asynchrony is handled in a different way altogether. We have something called Goroutines and here is no need for await statements in Go. Each of these Goroutines are started on their own threads that are lightweight (each of them have their own stacks, which defaults to 8KB in size) and the synchronization between each of them is achieved through communication through channels. These lightweight threads are capable of waiting asynchronously for any read operation to be performed on the channel and suspend themselves. The earlier implementation in Go is done using the SplitStacks technique. This implementation had its own problems as listed out here and replaced by Contigious Stacks. The article also talks about the newer implementation.
One important thing to note here is that it is not just the complexity involved in handling the continuation between the tasks that contribute to the approach chosen to implement Async/Await, there are other factors like Garbage Collection that play a role. GC process should be as performant as possible. If we move stacks around, GC becomes inefficient because accessing an object then would require thread synchronization.
Could you not implement Async-Await by simply memcopying the stack frames of async calls to/from heap?
In short, you can. As this answer states here, Chicken Scheme uses a something similar to what you are exploring. It begins by allocating everything on the stack and move the stack values to heap when it becomes too large for the GC activities (Chicken Scheme uses Generational GC). However, there are certain caveats with this kind of implementation. Take a look at this FAQ of Chicken Scheme. There is also lot of academic research in this area (linked in the answer referred to in the beginning of the paragraph, which I shall summarise under further readings) that you may want to look at.
Further Reading
Continuation Passing Style
call-with-current-continuation
The classic SICP book
This answer (contains few links to academic research in this area)
TLDR
The decision of which approach to be taken is subjective to factors that affect the overall usability and performance of the language. State Machines are not the only way to implement the Async/Await functionality as done in C# and Rust. Few languages like Go implement a Contigious Stack approach coordinated over channels for asynchronous operations. Chicken Scheme allocates everything on the stack and moves the recent stack value to heap in case it becomes heavy for its GC algorithm's performance. Moving stacks around has its own set of implications that affect garbage collection negatively. Going through the research done in this space will help you understand the advancements and rationale behind each of the approaches. At the same time, you should also give a thought to how you are planning on designing/implementing the other parts of your language for it be anywhere close to be usable in terms of performance and overall usability.
PS: Given the length of this answer, will be happy to correct any inconsistencies that may have crept in.
I have been looking into various strategies for doing this myseøf, because I naturally thi k I can design a language better than anybody else - same as you. I just want to emphasize that when I say better, I actually mean better as in tastes better for my liking, and not objectively better.
I have come to a few different approaches, and to summarize: It really depends on many other design choices you have made in the language.
It is all about compromises; each approach has advantages and disadvantages.
It feels like the compiler design community are still very focused on garbage collection and minimizing memory waste, and perhaps there is room for some innovation for more lazy and less purist language designers given the vast resources available to modern computers?
How about not having a call stack at all?
It is possible to implement a language without using a call stack.
Pass continuations. The function currently running is responsible for keeping and resuming the state of the caller. Async/await and generators come naturally.
Preallocated static memory addresses for all local variables in all declared functions in the entire program. This approach causes other problems, of course.
If this is your design, then asymc functions seem trivial
Tree shaped stack
With a tree shaped stack, you can keep all stack frames until the function is completely done. It does not matter if you allow progress on any ancestor stack frame, as long as you let the async frame live on until it is no longer needed.
Linear stack
How about serializing the function state? It seems like a variant of continuations.
Independent stack frames on the heap
Simply treat invocations like you treat other pointers to any value on the heap.
All of the above are trivialized approaches, but one thing they have in common related to your question:
Just find a way to store any locals needed to resume the function. And don't forget to store the program counter in the stack frame as well.
Could you provide some insight into the techniques that you use to ensure the quality of your solutions. For example, sometimes, I like to test my result using stopifnot() to ensure I'm not receiving ridiculous results. Are there any other techniques or functions that you use in data processing to ensure that you're receiving the solution you meant to?
Note: I realize that this is a broad question and perhaps a candidate for community wiki or even closure, but rather than voting to close, perhaps assist me by adding comments to direct the conversation.
Just a few things that come to mind (in random order)
This page has very interesting link for debugging in R (ok this is during production, but still related to your issue I think)
You can use exceptions, as explained in this discussions (and links therein)
You can write tests with known results (both for success and failure) and see that they actually do what they are supposed to do. Be sure to pass some weird data to the functions and see how they behave in a "not-so-normal" situation.
Don't just rely on automated tests: give your functions to a fairly computer illiterate person at work (not enough that he/she can't use R though!) and let him/her do some beta tests. You'll be amazed at the quantity of errors he/she will come up with!!! :)
Quality in software engineering is quite a massive area, and most of it applies to code written in R as much as code written in Cobol or C#, so my first answer would be 'it depends'.
For me, I come from the Pharmaceutical Industry, where what we do is regulated by government agencies like the FDA and the MHRA. For us, Quality is something we think about throughout the process so I would list the following as visible artifacts of quality;
We have a software development process, that's written down and repeatable (traditionally in this kind of industry this is a waterfall style, but more and more agile / prototyping style methodologies are being used)
We have a system that ensures every person involved knows what they should be doing (job descriptions) and is suitably qualified to do that job (training)
We start by defining what is required in some way, hopefully in some way that can be tested
We have some way of documenting our development process, where we've been and how (a combination of good documentation and Source Control)
We do testing wherever possible, and as early as possible (so, automated if possible)
We have people who are responsible for overseeing Quality, who are separate from people who are doing to prevent conflicts
We control the software environment that is used for development, testing and production (read; change control)
We control and manage software once it is in use, tracking issues and managing them (Issue Tracking)
We keep records, so that even if every person involved went under a bus / won the lottery the new people could still defend and prove everything above to a government inspector.
However, that's a big list, and I imagine their are lots of industries that don't do all of them (finance, education) and probably some who do more (building nuclear reactors, saving lives, NASA).
More specifically to what i assume you're getting at, before you code you should be able to define some specific starting input's and the answers you should get out, and I recommend you use something like RUnit or Testthat to build these in.
Closed. This question is opinion-based. It is not currently accepting answers.
Want to improve this question? Update the question so it can be answered with facts and citations by editing this post.
Closed 6 years ago.
Improve this question
Who is winning in the "Low vs High fidelity prototyping" debate?
Should prototype-zero (P0) be the first version of the final product? Or should be P-0 always a throwaway? What approach is the industry favoring?
Excelent article from wikipedia: Software prototyping
A prototype should always be a throwaway - a prototype is used to quickly prove a concept and influence the design of the real product. As such, a lot of things which are important for a real product (a thought-out architecture and design, reliability, security, maintainability, etc.) fall by the wayside. If you do take these things into account when building your prototype, you're not really building a prototype anymore.
My experience with prototypes where the code directly evolved into an actual product shows that the end-result suffers because of it - the lack of a real architecture resulted in a lot of cobbled-together code that had to be constantly hacked to add new features. I've even seen a case the original technology chosen for rapid development of the prototype was not the best choice for the actual product, and a complete re-write was necessary for V2.
I think we, the pedants, have lost this particular battle -- alleged "prototypes" (which by definition should be rewritten from scratch!!!-) are in fact being "evolved" into (often half-baked "betas"), etc.
Even today, I've applauded at the smart attempt by a colleague of mine to recapture the concept, even if the term is a lost battle: he's setting up a way for proofs of concept small projects to be developed (and, if the concept does get proven, transferred to software engineers for real prototyping, then development).
The idea is that, in our department, we have many people who aren't (and aren't in fact supposed to be!-) software developers, but are very smart, computer savvy, and in daily contact with the reality "in the trenches" -- they are the ones who are most likely to smell an opportunity for some potential innovation which could have real impact once implemented as a "production-ready" software project. Salespeople, account managers, business analysts, technology managers -- at our company, they all often fit this description.
But they're NOT going to program in C++, hardly at all in Java, maybe in Python but miles away from "productionized" -- indeed they're far more likely to whip up a smart proof of concept in php, javascript, perl, bash, Excel+VBA, and sundry other "quick and dirty" technologies we don't even want to dream about productionizing and supporting forevermore!-)
So by calling their prototypes "proofs of concept", we hope to encourage them to embody their daring concepts in concrete form (vague natural-language blabberings and much waving of hands being least useful, and alien to the company's culture anyway;-) and yet sharply indicate that such projects, if promoted to exist among the software engineers' goals and priorities, DO have to be programmed from scratch -- the proof-of-concept serves, at best, as a good draft/sketch spec for what the engineers are aiming for, definitely NOT to be incrementally enriched, but redone from the root up!-).
It's early to say how well this idea works -- ask me in three months, when we evaluate the quarter's endeavors (right now, we're just providing a blueprint for them, hot on the heels of evaluating last quarter's department- and company-wise undertakings!-).
Write the prototype, then keep refactoring it until it becomes the product.
The key is to not hesitate to refactor when necessary.
It helps to have few people working on it initially. With too many people working on something, refactoring becomes more difficult.
Response from BUNDALLAH, HAMISI
A prototype typically simulates only a few aspects of the features of the eventual program, and may be completely different from the eventual implementation.
Contrary to what my other colleagues have suggested above, I would NOT advise my boss to opt for the throw away prototype model. I am with Anita on this. Given the two prototype models and the circumstances provided, I would strongly advise the management (my boss) to opt for the evolutionary prototype model. The company being large with all the other variables given such as the complexity of the code, the newness of the programming language to be used, I would not use throw away prototype model. The throw away prototype model becomes the starting point from which users can re-examine their expectations and clarify their requirements. When this has been achieved, the prototype model is 'thrown away', and the system is formally developed based on the identified requirements (Crinnion, 1991). But with this situation, the users may not know all the requirements at once due to the complexity of the factors given in this particular situation. Evolutionary prototyping is the process of developing a computer system by a process of gradual refinement. Each refinement of the system contains a system specification and software development phase. In contrast to both the traditional waterfall approach and incremental prototyping, which required everyone to get everything right the first time this approach allows participants to reflect on lessons learned from the previous cycle(s). It is usual to go through three such cycles of gradual refinement. However there is nothing stopping a process of continual evolution which is often the case in many systems. According to Davis (1992), an evolutionary prototyping acknowledges that we do not understand all the requirements (as we have been told above that the system is complex, the company is large, the code will be complex, and the language is fairly new to the programming team). The main goal when using Evolutionary Prototyping is to build a very robust prototype in a structured manner and constantly refine it. The reason for this is that the Evolutionary prototype, when built, forms the heart of the new system, and the improvements and further requirements will be built. This technique allows the development team to add features, or make changes that couldn't be conceived during the requirements and design phase. For a system to be useful, it must evolve through use in its intended operational environment. A product is never "done;" it is always maturing as the usage environment change. Developers often try to define a system using their most familiar frame of reference--where they are currently (or rather, the current system status). They make assumptions about the way business will be conducted and the technology base on which the business will be implemented. A plan is enacted to develop the capability, and, sooner or later, something resembling the envisioned system is delivered. (SPC, 1997).
Evolutionary Prototypes have an advantage over Throwaway Prototypes in that they are functional systems. Although they may not have all the features the users have planned, they may be used on an interim basis until the final system is delivered.
In Evolutionary Prototyping, developers can focus themselves to develop parts of the system that they understand instead of working on developing a whole system. To minimize risk, the developer does not implement poorly understood features. The partial system is sent to customer sites. As users work with the system, they detect opportunities for new features and give requests for these features to developers. Developers then take these enhancement requests along with their own and use sound configuration-management practices to change the software-requirements specification, update the design, recode and retest. (Bersoff and Davis, 1991).
However, the main problems with evolutionary prototyping are due to poor management: Lack of defined milestones, lack of achievement - always putting off what would be in the present prototype until the next one, lack of proper evaluation, lack of clarity between a prototype and an implemented system, lack of continued commitment from users. This process requires a greater degree of sustained commitment from users for a longer time span than traditionally required. Users must be constantly informed as to what is going on and be completely aware of the expectations of the 'prototypes'.
References
Bersoff, E., Davis, A. (1991). Impacts of Life Cycle Models of Software Configuration Management. Comm. ACM.
Crinnion, J.(1991). Evolutionary Systems Development, a practical guide to the use of prototyping within a structured systems methodology. Plenum Press, New York.
Davis, A. (1992). Operational Prototyping: A new Development Approach. IEEE Software.
Software Productivity Consortium (SPC). (1997). Evolutionary Rapid Development. SPC document SPC-97057-CMC, version 01.00.04.
Locked. This question and its answers are locked because the question is off-topic but has historical significance. It is not currently accepting new answers or interactions.
How do you go about the requirements gathering phase? Does anyone have a good set of guidelines or tips to follow? What are some good questions to ask the stakeholders?
I am currently working on a new project and there are a lot of unknowns. I am in the process of coming up with a list of questions to ask the stakeholders. However I cant help but to feel that I am missing something or forgetting to ask a critical question.
You're almost certainly missing something. A lot of things, probably. Don't worry, it's ok. Even if you remembered everything and covered all the bases stakeholders aren't going to be able to give you very good, clear requirements without any point of reference. The best way to do this sort of thing is to get what you can from them now, then take that and give them something to react to. It can be a paper prototype, a mockup, version 0.1 of the software, whatever. Then they can start telling you what they really want.
See obligatory comic below...
In general, I try and get a feel for the business model my customer/client is trying to emulate with the application they want built. Are we building a glorified forms processor? Are we retrieving data from multiple sources in a single application to save time? Are we performing some kind of integration?
Once the general businesss model is established, I then move to the "must" and "must nots" for the application to dictate what data I can retrieve, who can perform what functions, etc.
Usually if you can get the customer to explain their model or workflow, you can move from there and find additional key questions.
The one question I always make sure to ask in some form or another is "What is the trickiest/most annoying thing you have to do when doing X. Typically the answer to that reveals the craziest business/data rule you'll have to implement.
Hope this helps!
Steve Yegge talks fun but there is money to be made in working out what other people's requirements are so i'd take his article with a pinch of salt.
Requirements gathering is incredibly tough because of the manner in which communication works. Its a four step process that is lossy in each step.
I have an idea in my head
I transform this into words and pictures
You interpret the pictures and words
You paint an image in your own mind of what my original idea was like
And humans fail miserably at this with worrying frequency through their adorable imperfections.
Agile does right in promoting iterative development. Getting early versions out to the client is important in identifying what features are most important (what ships in 0.1 - 0.5 ish), helps to keep you both on the right track in terms of how the application will work and quickly identifies the hidden features that you will miss.
The two main problem scenarios are the two ends of the scales:
Not having a freaking clue about what you are doing - get some domain experts
Having too many requirements - feature pit. - Question, cull (prioritise ;) ) features and use iterative development
Yegge does well in pointing out that domain experts are essential to produce good requirements because they know the business and have worked in it. They can help identify the core desire of the client and will help explain how their staff will use the system and what is important to the staff.
Alternatives and additions include trying to do the job yourself to get into the mindset or having a client staff member occasionally on-site, although the latter is unlikely to happen.
The feature pit is the other side, mostly full of failed government IT projects. Too much, too soon, not enough thought or application of realism (but what do you expect they have only about four years to make themselves feel important?). The aim here is to work out what the customer really wants.
As long as you work on getting the core components correct, efficient and bug-free clients usually remain tolerant of missing features that arrive in later shipments, as long as they eventually arrive. This is where iterative development really helps.
Remember to separate the client's ideas of what the program will be like and what they want the program to achieve.
Some clients can create confusion by communicating their requirements in the form of application features which may be poorly thought out or made redundant by much simpler functionality then they think they require. While I'm not advocating calling the client an idiot or not listening to them I feel that it is worth forever asking why they want a particular feature to get to its underlying purpose.
Remember that in either scenario it is of imperative importantance to root out the quickest path to fulfilling the customers core need and put you in a scenario where you are both profiting from the relationship.
Wow, where to start?
First, there is a set of knowledge someone should have to do analysis on some projects, but it really depends on what you are building for who. In other words, it makes a big difference if you are modifying an enterprise application for a Fortune 100 corporation, building an iPhone app, or adding functionality to a personal webpage.
Second, there are different kinds of requirements.
Objectives: What does the user want to accomplish?
Functional: What does the user need to do in order to reach their objective? (think steps to reach the objective/s)
Non-functional: What are the constraints your program needs to perform within? (think 10 vs 10k simultaneous users, growth, back-up, etc.)
Business rules: What dynamic constraints do you have to meet? (think calculations, definitions, legal concerns, etc.)
Third, the way to gather requirements most effectively, and then get feedback on them (which you will do, right?) is to use models. User cases and user stories are a model of what the user needs to do. Process models are another version of what needs to happen. System diagrams are just another model of how different parts of the program(s) interact. Good data modeling will define business concepts and show you the inputs, outputs, and changes that happen within your program. Models (and there are more than I listed) are really the key to the concern you list. A few good models will capture the needs and from models you can determine your requirements.
Fourth, get feedback. I know I mentioned this already, but you will not get everything right the first time, so get responses to what your customer wants.
As much as I appreciate requirements, and the models that drive them, users typically do not understand the ramifications of of all their requests. Constant communication with chances for review and feedback will give users a better understanding of what you are delivering. Further, they will refine their understanding based on what they see. Unless you're working for the government, iterations and / or prototypes are helpful.
First of all gather the requirements before you start coding. You can begin the design while you are gathering them depending on your project life cicle but you shouldn't ever start coding without them.
Requirements are a set of well written documents that protect both the client and yourself. Never forget that. If no requirement is present then it was not paid for (and thus it requires a formal change request), if it's present then it must be implemented and must work correctly.
Requirements must be testable. If a requirement cannot be tested then it isn't a requirement. That means something like, "The system "
Requirements must be concrete. That means stating "The system user interface shall be easy to use" is not a correct requirment.
In order to actually "gather" the requirements you need to first make sure you understand the businness model. The client will tell you what they want with its own words, it is your job to understand it and interpret it in the right context.
Make meetings with the client while you're developing the requirements. Describe them to the client with your own words and make sure you and the client have the same concept in the requirements.
Requirements require concise, testable example, but keep track of every other thing that comes up in the meetings, diagrams, doubts and try to mantain a record of every meeting.
If you can use an incremental life cycle, that will give you the ability to improve some bad gathered requirements.
You can never ask too many or "stupid" questions. The more questions you ask, the more answers you receive.
According to Steve Yegge that's the wrong question to ask. If you're gathering requirement it's already too late, your project is doomed.
High-level discussions about purpose, scope, limitations of operating environment, size, etc
Audition a single paragraph description of the system, hammer it out
Mock up UI
Formalize known requirements
Now iterate between 3 and 4 with more and more functional prototypes and more specs with more details. Write tests as you go. Do this until you have functional software and a complete, objective, testable requirements spec.
That's the dream. The reality is usually after a couple iterations everybody goes head-down and codes until there's a month left to test.
Gathering Business Requirements Are Bullshit - Steve Yegge
read the agile manifesto - working software is the only measurement for the success of a software project
get familiar with agile software practices - study Scrum , lean programming , xp etc - this will save you tremendous amount of time not only for the requirements gathering but also for the entire software development lifecycle
keep regular discussions with Customers and especially the future users and key-users
make sure you talk to the Persons understanding the problem domain - e.g. specialists in the field
Take small notes during the talks
After each CONVERSATION write an official requirement list and present it for approving. Later on it would be difficult to argue against all agreed documentation
make sure your Customers know approximately what are the approximate expenses in time and money for implementing "nice to have" requirements
make sure you label the requirements as "must have" , "should have" and "nice to have" from the very beginning, ensure Customers understand the differences between those types also
integrate all documents into the latest and final requirements analysis (or the current one for the iteration or whatever agile programming cycle you are using ... )
remember that requirements do change over the software life cycle , so gathering is one thing but managing and implementing another
KISS - keep it as simple as possible
study also the environment where the future system will reside - there are more and more technological restraints from legacy or surrounding systems , since the companies do not prefer to throw to the garbage the money they have invested for decades even if in our modern minds 20 years old code is garbage ...
Like most stages of the software development process its iteration works best.
First find out who your users are -- the XYZ dept,
Then find out where they fit into the organisation -- part of Z division,
Then find out what they do in general terms -- manage cash
Then in specific terms -- collect cash from tills, and check for till fraud.
Then you can start talking to them.
Ask what problem they want you want to solve -- you will get an answer like write a bamboozling system using OCR with shark technoligies.
Ignore that answer and ask some more questions to find out what the real problem is -- they cant read the till slips to reconcile the cash.
Agree a real solution with the users -- get a better ink ribbon supplier - or connect the electronic tills to the network and upload the logs to a central server.
Then agree in detail how they will measure the success of the project.
Then and only then propose and agree a detailed set of requirements.
I would suggest you to read Roger-Pressman's Software Engineering: A Practitioner's Approach
Before you go talking to the stakeholders/users/anyone be sure you will be able to put down the gathered information in a usefull and days-lasting way.
Use a sound-recorder if it is OK with the other person and the information is bulky.
If you heard something important and you need some reasonable time to write it down, you have two choices: ask the other person to wait a second, or say goodbye to that precious information. You wont remember it right, ask any neuro-scientist.
If you detect that a point need deeper review or that you need some document you just heard of, make sure you make a commitment with the other person to send that document or schedule another meeting with a more specific purpose. Never say "I'll remember to ask for that xls file" because in most cases you wont.
Not to long after the meeting, summarize all your notes, recordings and fresh thoughts. Just summarize it rigth. Create effective reminders for the commitments.
Again, just after the meeting, is the perfect time to understand why the gathering you just did was not as right as you thought at the end of the meeting. That's when you will be able to put down a lot of meaningful questions for another meeting.
I know the question was in the perspective of the pre-meeting, but please be aware that you can work on this matters before the meeting and end up with a much usefull, complete and quality gathering.
I've been using mind mapping (like a work breakdown structure) to help gather requirements and define the unknowns (the #1 project killer). Start at a high level and work your way down. You need to work with the sponsors, users and development team to ensure you get all the angles and don't miss anything. You can't be expected to know the entire scope of what they want without their involvement...you - as a project manager/BA - need to get them involved (most important part of the job).
There are some great ideas here already. Here are some requirements gathering principles that I always like to keep in mind:
Know the difference between the user and the customer.
The business owners that approve the shiny project are usually the customers. However, a devastating mistake is the tendency to confuse them as the user. The customer is usually the person that recognizes the need for your product, but the user is the person that will actually be using the solution (and will most likely complain later about a requirement your product did not meet).
Go to more than one person
Because we’re all human, and we tend to not remember every excruciating detail. You increase your likelihood of finding missed requirements as you talk to more people and cross-check.
Avoid specials
When a user asks for something very specific, be wary. Always question the biases and see if this will really make your product better.
Prototype
Don’t wait till launch to show what you have to the user. Do frequent prototypes (you can even call them beta versions) and get constant feedback throughout the development process. You’ll probably find more requirements as you do this.
I recently started using the concepts, standards and templates defined by the International Institute of Business Analysts organization (IIBA).
They have a pretty good BOK (Book of Knowledge) that can be downloaded from their website. They do also have a certificate.
Requirements Engineering is a bit of an art, there are lots of different ways to go about it, you really have to tailor it to your project and the stakeholders involved. A good place to start is with Requirements Engineering by Karl Wiegers:
http://www.amazon.com/Software-Requirements-Second-Pro-Best-Practices/dp/0735618798/ref=pd_bbs_sr_2?ie=UTF8&s=books&qid=1234910330&sr=8-2
and a requirements engineering process which may consist of a number of steps e.g.:
Elicitation - for the basis for discussion with the business
Analysis and Description - a technical description for the purpose of the developers
Elaboration, Clarification, Verification and Negotiation - further refinement of the requirements
Also, there are a number of ways of documenting the requirements (Use Cases, Prototypes, Specifications, Modelling Languages). Each have their advantages and disadvantages. For example prototypes are very good for elicitation of ideas from the business and discussion of ideas.
I generally find that writing a set of use cases and including wireframe prototypes works well to identify an initial set of requirements. From that point it's a continual process of working with technical people and business people to further clarify and elaborate on the requirements. Keeping track of what was initially agreed and tracking additional requirements are essential to avoid scope creep. Negotiation plays a bit part here also between the various parties as per the Broken Iron Triangle (http://www.ambysoft.com/essays/brokenTriangle.html).
IMO the most important first step is to set up a dictornary of domain-specific words. When your client says "order", what does he mean? Something he receives from his customers or something he sends to his suppliers? Or maybe both?
Find the keywords in the stakeholders' business, and let them explain those words until you comprehend their meaning in the process. Without that, you will have a hard time trying to understand the requirements.
i wrote a blog article about the approach i use:
http://pm4web.blogspot.com/2008/10/needs-analysis-for-business-websites.html
basically: questions to ask your client before building their website.
i should add this questionnaire sheet is only geared towards basic website builds - like a business web presence. totally different story if you are talking about web-based software. although some of it is still relavant (e.g. questions relating to look and feel).
LM
I prefer to keep my requirements gathering process as simple, direct and thorough as possible. You can download a sample document that I use as a template for my projects at this blog posting: http://allthingscs.blogspot.com/2011/03/documenting-software-architectural.html