UML Fever

CowboyRobot writes "Queue has a couple of articles about UML: Death by UML Fever by Boeing software architect Alex Bell describes the problems that can result from over-reliance on modeling tools, with lighthearted lessons for the software development process in general and numerous illuminating quotations, such as: "Good judgment comes from experience. Experience comes from bad judgment. - Jim Horning." Then, one of the developers of UML, Grady Booch of IBM, follows with The Fever is Real, in which he explains the motivations for creating the language, how it's used today, and where he expects it to go soon."

Comments

Acronyms

[Space cowboy]

True linux-focussed geeks would have immediately wondered why user-mode linux was suddenly such a hot topic, and so dangerous to boot :-)

Simon

Re:Acronyms

[dotz]

If you don't know all the meanings of UML acronym, you can't be "true" and "geek" at the same time.

Am I the only person, that finds all those oh-so-funny usermode-linux jokes really not-so-funny?

Re: Everything, including tools, in moderation!

[Face the Facts]

Design patterns and UML were designed as practical tools, not dogma. If they help you do what you were doing anyway (and they often do), then great: use 'em. But if they don't, then don't. They're there to serve you, not the other way around.

This is true, but note that the UML/patterns/OO newbie is in no position to determine that. One common mistake is to read the book, discard the parts you don't think is necessary, and then proceed with your design work. The rules that you chose to ignore were put there by pretty smart people, and there's a good chance they were put there for a good reason. When the design finally fails because you were missing something, the egotistical designer then blames the method.

The point is, I think the parent post was suggesting that the programmers in question may simply have broken the rules, and not actually found some instance where the methods really apply poorly. It's ego-boosting to think that what you do is unique and beyond the reach of old stuffy rules, but the truth is that most of us are doing things that have been done before.

This isn't to say that those cases don't exist, but that they're probably rarer than you think, especially if your team of programmers is trying it out for the first time, especially if you don't have a senior engineer already experienced in the method guiding your team. For the first time, at least, the instructions should be followed to the letter and strictly enforced. They should be dogma until you've at least went through a complete product life cycle with them.

What you suggest we've already tried for decades. The result is prevasively poor documentation and fragile designs.

Software architect?

[AtariAmarok]

Never forget Weinberg's Law:

"If builders built buildings the way programmers wrote programs, then the first woodpecker that came along would destroy civilization"

Re:Software architect?

[spektr]

"If builders built buildings the way programmers wrote programs, then the first woodpecker that came along would destroy civilization"

The leading building corporation would proclaim that there's nothing wrong with the buildings and a new market of woodpecker traps and anti woodpecker missiles would thrive.

Re:Software architect?

[His name cannot be s]

Aye, laddie.

Be thankful of the fact we've (as humans) have been building shelter for 20000+ years.

As software developers, we've been making stuff for a little over 50 years.

We are In the MudHut age of Software Development:

Garrett's Rule #3: "MudHut software" We are continually re-writing software that performs the functionality we require, simply to make the software run on the platform that we rewrote, so that we can make better software.

You see, 20,000 years ago, our clever, but inexperienced ancestors had to make new shelter every few days/months/years/whatever, simply because they made them from mud.

When the next wind, rain or hail came, it outdated their technology, and forced them to upgrade.

took thousands of years to make the difference, an build structures that could stand up to the changing environment.

So, just when it looks like MS will rule the desktop forever, just remember, while some clever frickin' caveman probably added the first straw and dung to his hut, and improved it, it never lasted. Cooler heads prevailed.

Now, for those who didn't catch it: I just compared MS products to shit and straw.

Thanks,

Re:Software architect?

[skaffen42]

Tacoma Narrows anybody?

I work with a couple of civil engineers and in a previous life I did some IT work at a construction company. What this has taught me is that Weinberg's Law was thought up by people who never stepped onto a construction site.

Yes, there is a lot more formalized rules and many more years of experience in the construction industry, but people are still people and they get lazy and cut corners just like in any industry.

I guess what I am trying to say is that we can learn a lot from the brick and mortar crowd, but don't think they have all the answers.

Re:Software architect?

[shakah]

...but people are still people and they get lazy and cut corners just like in any industry.

I heard/read a quote from someone to the effect that "anyone can build a bridge that will stand for 100 years -- it takes engineering to build one that will just barely stand for 100 years", with the point being that engineering consists of balancing/estimating/reconciling competing constraints (e.g. budgets, usage patterns, loads, schedules, etc.).

Was the Tacoma Narrows collapse due to cutting corners? I thought it was more related to ignorance of how winds might effect a bridge as opposed to what might be thought of as lousy engineering. And at least civil engineers learned from the failure -- isn't wind tunnel testing a standard practice in bridge design these days?

Re:Software architect?

[Kevin Stevens]

I often hear building type analogies like this, and IMHO always focus on the wrong aspects of software and architecture.

The reason most software is late, over budget, and buggy is because the specs are not what the customer wants. Customers understand that when they are buildng a house, that the blueprints are of utmost importance, and that the blueprints are going to reflect the house that they are going to build, and go to great lengths to visualize the house depicted in the blueprints. They also have an understanding of building concepts, and know that they can not wait until the bathroom is finished, and then say "hey, I want the tub over on the other side of the room" without understanding that the bathroom will essentially have to be gutted. These same people will have no problem demanding significant overhauls of the layout of a gui or the basic fundamental requirements of the software. From my experience, customers will often blow right through a spec document, browsing it, and saying OK looks great! What is worse is, the more detailed (hence boring) the spec is, the less likely the customer is to really look at it. Begging, pleading, blackmailing, and torturing customers to read the specs is rarely effective, the best solution IMHO is the XP method of letting the customers see things as early as possible.

This leads to another problem- software growing in scope and function well beyond its original intentions. People inherently understand that you can not take a basic mudhut or woodframe house and just keep adding to it indefinitely and still have a coherent structure- there is only so much the foundation can support, and only so much you can do to a house before you have to major extensive rennovations. The same thing occurs with software. Customers often say, "we like this, but we would really like this feature," and management will almost always cave in and say sure! Next thing you know, the system has a large patchwork of add-ons that sometimes have to be hacked in and make the code difficult to maintain, and make it even more rigid. People dont take trailers or sheds and try to make them into colonial two story homes, but they will often try to take simple reporting tools and make them into comprehensive organization wide information dashboards, one report or feature at a time.

Design Patterns and frameworks try to attack the problem by telling us how to build even sheds with concrete and steel foundations, and modularized/interchangable parts so that further additions do not cause too many problems, which is a good thing. But until there is a universal general understanding of the software development process by the customers, these problems will persist in software, and the industry itself.

Re: Everything, including tools, in moderation!

[mwheeler01]

I don't think his point is that projects are too unique to be guided by such rules. He's merely saying that design patterns are often useful, but don't make sacrifices in your own design to adhere to specific pattern rules. It would be foolhardy to think that most problems could be solved by using straight from the text patterns.

Also I think he's driving at another point: you should only model as much as you need to. I think modeling is useful to help everyone understand how the system works but it's not always necessary to model every behavior before you continue with the project.

UML

[JaxWeb]

Who uses UML? The designers or the programmers?

I should imagine the designers wouldn't be very good at it, and I should imagine that programmers would have better ways to express themselves.

Sorry, this has just been a question I've wanted to know the answer to. And the story has been /.'ed.

Re:UML

[mwheeler01]

Who uses UML? The designers or the programmers?

I should imagine the designers wouldn't be very good at it, and I should imagine that programmers would have better ways to express themselves.

Actually a good designer probably has some experience coding. As to who uses it? Well ideally the designer or project leader should author it and the programmers should be able to read and understand it, so both parties technically use it.

Re:UML

[BigGerman]

I could not get the article either and there were only 7 slashdot comments...
I personally (speaking from 12 years of experience) found UML useful in many situations when certain (already coded) solution was presented to someone else. So just to illustrate some point, it is much easier to sketch a few diagrams than to write a textual description.
On the other hand I have seen environments where the UML was adopted as sort of cornerstone of top-down approach: all the design is done first and captured in UML. The problem is that disagrams become absolete very quickly and people stop referring to them which accelerates the decay even more.

Re:UML

[sporty]

As a programmer, who has used various UML tools, doing OOD (oo development) is a little easier with a graphical tool. It becomes REALLY easy to see how everything relates. Think of it like an ERD (entity relationship diagram) for a DB schema. Once I have my class diagrams and what not, i can generate java from them. I don't have to type out the class framework. Better yet, I can take an existing project and show how everything relates and in turn, have a map of an existing project.

Re:UML

[Woogiemonger]

I've rarely seen where the designers aren't the programmers as well, at least somewhat. If there are people doing the design for a software, they're generally the lead developers and have a better chance at coming up with an intelligible, well-crafted set of UML diagrams than the programmers not involved in software architecting. When you say "better ways to express themselves", what ways are you referring to? While UML can be considered overkill, it certainly covers the bases.

Re:UML

[Oligonicella]

Been designing and implementing systems for 35 years. I use portions of UML. I find chunks of it equivalent to blithering. Others are excellent ways to communicate to clients. Those parts happen to have existed since the 70-80's. Not much of use is really new here.

I don't really see a significant distinction between "designers" and "programmers". One needs to know each to do well.

Re:UML

[jeffy124]

if you've ever worked on a program by simply drawing boxes of components and drawing lines between them, you've engaged in modeling. Or even some simple flowcharting is also modeling. UML is a formalized method to stuff like that (your component diagram would be referred to as a structural model, the flowchart a behavioral model).

Who uses UML? The designers or the programmers?

Almost everyone in a development project. Designers, programmers, testers, etc. Even management and your customer can understand what's going on in most of the models.

As for who would be better, I dont know yet. I'll know better in about a year, as we've started a few UML/RUP projects where I work.

Re:UML

[toesate]

Before reading on, let me just say that I am a proponent of using UML/IDL/Design patterns. So..

One of the idea of UML is that this framework allows for people of different roles to converse in a common modelling language.

A business/product person will easily understand Use Case and high level sequence diagrams.

A product architect/designer is able to transform these use cases and sequence diagrams into "lower level requirements" - more sequence diagrams, state diagrams, and define the IDL interfaces relationships between interacting "boxes".

An engineer is able to take in these low level requirements, read the state diagrams and IDLs, create the component and class diagrams, and generate the skeleton codes for actual implementation.

From managing requirements point of view, UML is a good tool, provided that everyone in the food chain understand their roles and know how to create proper diagrams and read their counterpart's diagrams. (You cannot blame the tool if someone do not know how to do their work.)

In case someone misunderstood me, UML itself is not sufficient for software development, but with UML, the whole software development can be a much pleasant tasks.

Overall, (I think) UML does make my engineering team more efficient - you know when the first thing my fellow engineers do after opening a specification document is to look for some UML diagrams.

Re:UML

[a!b!c!]

The main people who use UML are Consultant Software Architects, who come into a project. Create a solution, and provide the UML framework, and collect a big fat pay check. They then move on the next project.

Meanwhile, the programmers who are stuck with this static UML framework, while the timeline, user and system requirements are changing, are put in a precarious situation where the design just doesn't work anymore, and often the project crumbles.

Back to the Architects, who claim there design was adequate for the problem, and they don't understand how the project failed. Must be the programmers.

Re:UML

[toesate]

If the programmer and designer aren't the same person, I'd argue that the project is already in trouble.

Not necessarily, I'd argue otherwise.

One of the way to make this work is to bring Xtreme Programming from coding to designing too.

For example, you design this side of the interface, I design the other side of the interface. And then, I implement this side of the interface that you had design, and you implement the other side of the interface that I had design.

So, you see, the designer and programmer are different persons, one design and the other code what was designed. And this can work pretty well actually.

Of course, for argument sake, this can still means that the programmer and the designer is the same person, since they are both working on the /same bigger thing/.

This also requires both parties (you and I in this case) to trash out any misunderstanding during designing phase, or during early stages of coding. And it will definitely help when both are versed in, for example UML.

Re:UML

[plover]

Your post reminded me of my initial reluctance to adopt UML techniques. My first exposure to this was about 10 years ago when Booch, Rumbaugh, et al, were still busy duking it out whether you should draw relationships as dashed lines or dotted lines to red clouds or black boxes. There were some really heinous proposals floating around (Booch had some of the worst, as I recall, because they all centered on "you should draw this thing exactly this way," where a stray curlique on the left side of a comment box might mean "but only in odd years preceding a leap year.")

UML still suffers from some of that: filled arrowheads vs. stick arrowheads, for example. Arrows do inherently express a relationship (at least to me,) but neither filled nor hollow represents "synchronous" vs "asynchronous". The UML is an 800 page standard, of which about 750 pages are useless detail fluff of this nature. Yet they're all "part of the standard" and thus are all equally important, at least to the AR+ people who religiously push these things. I understand that if you want to precisely express an idea that you should use the precise syntax in the language. But practical use of UML shouldn't require knowing all 800 pages, just as writing practical C++ code doesn't require an intimate knowledge of templates.

Please don't dismiss me immediately as a UML critic (or a template critic!) One of the places that I find it of value is that it gives us developers a common ground on which to meet the analysts and business users. But much of the syntax is simply too cryptic to be of real working value in expressing ideas. And if you look at developers as the "compilers" of UML (the people who have to interpret it into code) we would have to know every last comma of the standard to perfectly implement it. Not a realistic assessment.

Designs cast in Stone

[Titusdot Groan]

My first encounter with a group using UML was when I was called in to review the design of a medical imaging system. I learned UML, thought it was pretty cool, and then reviewed the design.

I found a significant design flaw with a pattern that they were using everywhere and raised it as an issue. Turns out they knew about the flaw but didn't want to fix it since they would have to redo all their UML diagrams for the bulk of the project.

Then they had the nerve to ask me to signoff on the design. Further they had the nerve to ask me to use UML on my next project!

UML is really great for communication of designs. But if you invest too much into the UML you end up "coding" the proposed design and it becomes too hard to modify.

Where UML really shines is in HUGE projects, where systems analysts design the system in UML and programmers write the code. Anything else it's a little bit of overkill and a little dangerous.

Re:Designs cast in Stone

[naden]

Parent is exactly right:

UML is perfect for HUGE projects where you have a large number of developers, designers and managers all of whom must understand what is going to be delivered.

Everything smaller should be interface driven. That is, you do storyboards and interface mockups as the next step after the requirements specs. THEN you give those mockups to the programmers and tell them to implement it.

This tends to result in the best quality programs because it comes from a user-centred point of view rather than a programming point of view.

This is how you explain the difference between OSX (top down) versus Linux (bottom up) approach to developing the OS.

Re:Designs cast in Stone

[rSelrahc]

Actually, UML can also be useful for smaller projects, but used in a different way...

For smaller projects, you can use UML in a UMLAsSketch mode (http://www.martinfowler.com/bliki/UmlAsSketch.htm l). In this mode, all UML does is to allow for a common graphical "language" for expressing your ideas.

For larger project, you can move on to UMLAsBlueprint (http://www.martinfowler.com/bliki/UmlAsBlueprint. html).

For certain, very specific projects, with very well defined semantics, and very well defined transformations, you can move on to using UMLAsProgramminglanguage (http://www.martinfowler.com/bliki/UmlAsProgrammin gLanguage.html).

Note that I do not completely agree with where Fowler puts his markers in the continuum of the modes to using UML. I think, for example, that his UMLAsBlueprint is too close to his UMLAsProgramminglanguage. But this may be due to his propensity towards Agile methods.

One thing to keep in mind is that UML is a "language"! It is a means to express a software system. It is _NOT_ a panacea to all software development woes and it does _NOT_ replace good architecture practices! Anyone who thinks the opposite is doomed to failure (and anyone who improvises themselves software architect because they know UML doubly so!).

Re:Designs cast in Stone

[bluGill]

With a good UML tool it should be very easy to make changes. UML is worthless if you are not making changes to the models. Indeed if your UML is set in stone you should throw those stone tablets away.

UML used correctly is a good tool for getting your design down. Print that design poster size and hang it on your wall so you can at a glance see the design! However from several years of using UML I have discovered that you re-print that poster at least once a week because it goes out of date that quickly, and last weeks poster is useless. (even though that change is minor you can assume your work for the week will be based on the parts you changed)

Booch - there was nothing before he came along?

[Anonymous Coward]

Contrary to his opinion, the one he never ceases to state, over and over again, there were modelling systems before he 'invented' them all. There have always been ways to annotate a software design with clarity and there always will be.
Has anyone ever worked with anyone who is totally focussed on the UML? We've had to get rid of them as they don't contribute anything except endless debates and meetings that are way too long. Sometimes you have to deliver stuff. I bet he missed that bit while he was writing papers about how revolutionary he is.

Re: Everything, including tools, in moderation!

[rSelrahc]

Let's not forget that a pattern is a solution to a problem in a context. If the context does not match your situation, then don't use it - even if the problem is the same and the solution is attractive.

Ground Realities

[frodo from middle ea]

All the so called goodies of UML asides, the fact is lot of projects have a very tight schedule and overlook lot of potential pitfalls.

When these overlooked potential pitfalls, start to pust the project dates further and further away, the entire s/w development cycle is compromised and lot of design work is bypassed. Most of the functionality is directly coded from the requirement specs, bypassing all the UML stuff.

And there is the issue of Change management, Over the lifetime of a project , lot of staff changes including developers and designers, and the documentation and UML stuff starts to lag behind the current implementations. There comes a time when the UML diagrams represention is no where near the current functionality of the Code.

Having worked on two very long maintainace projects for very rich clients , I can tell you that initial project delivery is only the tip of the ice burgh. The real deal is maintaining and upgrading the project over a long time, and sadly UML is very inadequate for that.

From a developer's POV , who needs to work on someone else's code ,Nothing and really Nothing is as useful as proper comments in a code.

Re:Ground Realities

[Malc]

"Most of the functionality is directly coded from the requirement specs, bypassing all the UML stuff."

In my experience that means the requirements have changed (feature creep, which kills many projects), or the implementation design wasn't done properly in the first place. You have to do the thinking at some point, and I would rather do as much as possible before I'm stuck in to the code. The code becomes easier to write, and the changes to the architecture are easier to make. Short sweet prototypes during a design phase are critical to help scope out unknown and risky parts of the implementation.

Boring, but true story

[smallfeet]

A worked on a new development project a while back and we decided to try XP for the design and development cycle. Another project in the same department started at about the same time and used Rational Rose and produced a lot of UML design specs up front. We had part of our application up and running to the users satisfaction within 3 months, but then ran into a major design oversight that bogged us down for the next 3 months. The other project didn't start to program for 2 months and didn't have anything really to show the customer after 6 months. In the end both projects were killed.

The moral: There are no magic bullets.

Re:Boring, but true story

[toesate]

In the end both projects were killed.

This is unfortunate..

There are many reason why projects fail, so I do not want to speculate. But my favourite is Anti-Patterns in Project Management (review)

If it is a major design flaw discovered late, then this really sucks, especially if it is due to lack of foresights or experiences.. But hey, that is how we gain experiences.

Actually my experience tells me that UML can be very suitable for fast prototyping. And one of the lethal combination for success, in my opinion is UML+IDL+CORBA(Java)+eXtreme Programming.

It can be quite amazing how a prototyping team can churned out results, with the help of everything good, like generated skeletons, and stubs, and it helped us to get something very functional up and running, and then allows for subsequent iterative improvements that is almost transparent to the users, even changes implementation between C++ and Java. Of course, not forgetting dedicated engineers and everyone's commitment.

UML and OO

[No. 24601]

While UML is better suited to OO designs, it does provide facilities (activity diagrams) for designing algorithms in say C or Pascal and for describing system interaction with external components (sequence diagrams). However, UML is particularly directed at OO development and using it to develop software in non-OO languages requires a bit of a hack.

One facility I think is particularly absent from UML is modelling of error conditions. In most cases it's awkward to indicate that an error might exist and how to handle it if does occur. Also, UML notation seems incredibly basic (gulp!) and simply can't tackle algorithmic modelling for complex applications like memory management or software interrupt handling in an operating system.

If not already underway, the open source community should step forward and come up with a more advanced alternative to UML and then perhaps demonstrate its usefulness by applying it to a complete modelling of say.. oh.. umm... Linux ;)

I've gotten spams about these

[AtariAmarok]

The leading building corporation would proclaim that there's nothing wrong with the buildings and a new market of woodpecker traps and anti woodpecker missiles would thrive.

I've already gotten spams about these products concerning software woodpeckers. In fact, I got one this morning that had a title "SOLVE YOUR SOFT PECKER PROBLEMS"

Design is a Good Thing.

[Millennium]

UML and other design tools are Good Things, and they should be used more often. They really do tend to make things easier in the long run.

However, they do have a major weakness: it's possible to become too reliant on design documents, such that one loses the ability to think on one's feet. It's an easy trap to fall into, but it has to be avoided. Otherwise, any design problem -and such problems are inevitable in any project of significant size- become paralyzing.

Re:Design is a Good Thing.

[enjo13]

I have the same issue with front-loaded design that I have with heavily commenting code.. it's just one more thing to maintain.

UML and various other design schemes have this funny habit of making you code every program twice. Once in a modelling language and once in actual code.. A change in requirements, methodology, bug fixes, or anything else that causes code modification requires you to not only update your code but your design diagrams as well. Throw in a heavy handed comment scheme (the comments have to be updated as well!) and it's really a maintenence nightmare.

Instead I prefer high level documentation and diagramming. Instead of designing at the lowest levels of a system I have a small number of very high level diagrams and documents that describe the overall system. I then try to write very readable code that allows someone with a basic understanding of the system (from the documents) to easily find their way in the code and begin to actually USE it at the lowest levels.

Programmers don't build programs

[wowbagger]

This analogy is flawed.

Programmers are NOT to programs what builders are to buildings.

Programmers are to programs what ARCHITECTS are to buildings.

Builders are to buildings what COMPILERS AND MAKE UTILITIES are to programs.

Now, I suggest that you make an architect work under the same constraints as a programmer:

1) I cannot tell you were the house will be built, so you cannot estimate heating/cooling, snow loads, etc. Can't it figure it out itself?
2) I cannot tell you that the house won't be moved to a completely different climate once built - can't you make it automatically adjust?
3) I cannot tell you what building materials will be available to build the house with. Can't you make your design work just as well with wood as with adobe?
4) I cannot tell you how many rooms will be needed. Can't the house automatically add rooms as needed?
5) I cannot tell you what services, such as electricity, water, and gas, will be available. Can't you make the house work just as well on wind power as grid power?
6) I can tell you that whatever I told you is subject to change without notice.
7) Oh, by the way, now that the house is almost designed - Add a hanger. No, I will NOT tell you whether the hanger is for a Cessna 182 or a 747 - can't you make the hanger figure that out when I park the plane?
8) Oh, the house cannot cost more than US$10,000.
9) Oh, and the house must be build on a quarter-acre lot.
10) Oh, and the house must be ready to move in tomorrow. Morning. Before I go to work.
11) Oh, and the house must be built by two dain-bramaged monkeys with Nerf Tools.
12) Did I mention being earthquake-proof?
13) Oh, the lot is in a floodplain. Can you make it water-tight? Or float? Or both?
14) Hey, how about a houseboat?

Re:Programmers don't build programs

[master_p]

In other words, the problem is not having a concrete set of requirements to begin with. Which means the problem lies with the management, and not with the programmers. If the management decides that its software requirements must be cast in stone, then the programmers are happy. If the management keeps changing requirements every morning, then we have all the usual problems.

Re:Programmers don't build programs

[WayneConrad]

In other words, the problem is not having a concrete set of requirements to begin with.

It's worse than that. The problem is that, for all but the most trivial of problems, you can't have a concrete set of requirements.

If you could have a concrete set of requirements that would unambiguously result in code that met the customer's needs, then you could automate the process of turning the requirements into code. In other words, the requirements would be code.

As much as management would love to be able to automate the process of turning requirments into code, they can't. Nor can they even create a process that someone can follow to turn requirements into code (if you could, you could automate following the process and not need the programmer).

The bottom line is, you can't ever have a complete set of requirements. The best thing to do is to get over it and not work under fictional processes that assume you can ("If only we could get the requirements frozen, then we could ship on time"). Requirements change because writing code teaches the programmers and the business that the requirements were unimplementable in the time required, or that they don't actually meet the needs of the business. That's ok. It better be, because that's reality.

Re:Programmers don't build programs

[Bozdune]

That's certainly true, but the "usual problems" can be mitigated with careful thought and good design. I've built lots of software from inadequate specifications. It's possible to do a good job if you take the time to understand the problem domain, even if the specification is poor or nonexistent.

Neither of the following examples involved me, but I was an interested contemporaneous observer of both.

Case study #1: Web application (buying portal). The app was designed with J2EE tools, and the Architect allowed the tools to automagically generate the schema that supported the app. Problem: The Architect did not consider that reports would need to be generated from the schema, some day. When it came time to write them, the task was horrific and nearly impossible, and the team almost failed. Who is guilty, management, for not specifying all the reports up front, or the architect, for not anticipating the need for reporting? The Architect, of course, blamed management; management, naturally, blamed the Architect.

Case study #2: A very arge commercial bank wanted a system for loan portfolio analysis. The Architect endured management displeasure over his long-ish timeline by designing a pseudo-language first, in which the team ultimately wrote the portfolio analysis package. The Architect did this because he envisioned massive downstream changes to the specification. Numerous changes to the specification then occurred over a two-year period, just as the Architect had predicted. Result: The Architect's company pocketed huge sums by bidding half of what the competition bid on the same project, and incurring almost zero cost, because the pseudo-language changes were so trivial to make.

From experience...

[IA-Outdoors]

I'm sure what I am about to say is covered in part by previous posts but I wanted to try and articulate it a little better. First, before there was UML Object Oriented Analysis (OOA) and Design (OOD) existed. I'm stating the obvious for some but this is often forgot. UML is just a standard language for doing OOA and OOD. UML does not tell you *how* to do the analysis or the design and, in fact, programmers that are perfectly capable of reading and implementing things represented in UML are unable to do the analysis and design needed to build the UML.

So, as a software manager, when I hear an interviewee answer the question "What do you know about UML" and their answer is something like "I have done it using Rational's tools" I want to puke. UML automation tools are nothing more than long rope to hang yourself with when given to a programmer who knows nothing about OOA and OOD.

My advice is to learn OOA and OOD independently from UML and then when you have a full grasp of that to look at how UML might help you in those efforts.

Two Words: Sprint

[That's Mister Jesus]

OK, that's not really two words, but two words with a colon after it is much funnier than one word with a colon after it. Anyway, folks from the Kansas City area are familiar with Sprint, for whom we've all worked occassionally. It's your typical environment where management is non-technical but deluded to the contrary and they overcommit to things like RUP, UML, and code generation toold. Some people will tell you there should be sequence diagrams for every single bit of logic in an application and so on. It's all been said before and I'm sure there are many Sprint's out there. The applications work eventually, but at great cost, delay and frustration.

In my rarely humble opinion, UML is misused as just another management crutch, like powerpoint and outsourcing. Managers want to wow their uppers with charts and graphs and a few simple innovative ideas rather than than be hands on and make the tough decisions required to bring a project in on time and on budget. They put all their faith in an ideas like UML and RUP that they themselves don't fully comprehend and expect the world to magically change.

I was recently in a working group meeting for an insurance standards body called Acord. Somebody whipped out an interaction diagram and all these MBA's thought it was the second coming of Christ(or the first coming if you're Jewish).

Re: Everything, including tools, in moderation!

[mwood]

The way I read that quote, the idea is not to pick and choose the rules you like, but rather to throw out the whole thing if it is not working for you. If pounding nails in with the butt of a screwdriver is not getting the job done right, don't put a different handle on the screwdriver; get a hammer.

I see here a reaction to something I'd noted myself: for a number of developers, OO ceases to be a tool in the toolbag and becomes their religion -- they want to evangelize all of the unconverted, and strive to see every aspect of life through the OO lens. The book I'm reading right now contains some side-splitting examples of using the OO sledgehammer to crack a programming peanut, and I'm sorry to say that in my experience this practice is not confined to tutorial material.

There's a step that comes before "what's the best way to code this?" It's "what's the best way to *design* this?"

Re:UML Modelling - Communications Gap

[gbjbaanb]

don't forget that UML usage is driven mainly by:

a) the UML tools vendors who say its the best way to produce quality in your projects
b) the people (managers usually) who believe all the stuff a) wrote.

Personally, after seeing UML used to get nowhere, I would always go for a lightweight development methodology (like XP which I dislike, or Crystal Clear)

Is UML really a language?

[pohl]

What is the justification for the 'L' in 'UML'. I realize that in the loosest sense, UML is intended to be a means of communicating ideas, but it feels extremely primitive, like cave drawings. There's not even a standard machine-parseable representation for UML so that various programs can generate/manipulate/transform UML models without reverse-engineering some proprietary cryptic file format. And as for communicating person-to-person, I find it much more practical to use design-pattern language and a few terms from The Jargon File to improve communication with my team. Think about the jargon used between surgeons and their assistants -- to me, that's the kind of language that actually improves communication where it matters...UML seem to take the part of the development cycle that proceeds at the most frustratingly glacial pace and make it even slower. And that seems to make it different from all of the other languages that I've learned.

Re:Is UML really a language?

[rSelrahc]

Things are changing... Look at the Eclipse UML2 project (http://www.eclipse.org/uml2/).

There is an OMG standard for the serialization of UML model: XMI. And there is an OMG guideline on transforming the models: MDA.

Now, if only all tools adhered to the XMI standard as their main storage method... Although most can currently output some variant of XMI (although there's just too many right now).

But then again, this is all about tools. There is nothing that would prevent you from using the UML notation on a whiteboard...

Don't knock mud 'till you've tried it!

[Chemisor]

> You see, 20,000 years ago, our clever, but
> inexperienced ancestors had to make new shelter
> every few days/months/years/whatever, simply
> because they made them from mud.

Buildings made of mud (in adobe brick or rammed earth construction, as it is called these days) are much sturdier than the frame matchbox you are living in. In fact, rammed earth walls are so durable that some of them are still standing after 5000 years of rain/wind/hail etc. And they are more energy efficient too.

Re:Don't knock mud 'till you've tried it!

[AJWM]

In fact, rammed earth walls are so durable that some of them are still standing after 5000 years of rain/wind/hail etc. (emphasis added)

What you don't see after 5000 years is all the hundreds of thousands of mud (rammed earth, etc) buildings that didn't stand up to five millenia of rain/wind/hail etc. There are some mud hills in the middle east that were pretty amazing structures in their day.

I've heard similar marvelling over how the stone work of the Incas has amazingly survived some pretty severe earthquakes. Yes, the stuff that's still standing is pretty darn impressive (I've seen quite a bit of it first hand), what you don't see or hear about is the rubble where some of that stonework didn't stand up to the quakes.

Don't judge any technology by its successes unless you also know of its failures.

Just use common sense!

[mrtom852]

How many people use ad-hoc diagrams? Most of the best diagrams I've seen don't follow any standard but do a better job of documenting.

Deployment diagrams are the best example I can think of as they are the most common to not be done in the UML yet they are always easy to understand. (I never could figure out how to describe clustered/redundant/distributed environments with the UML)

I think class and sequence diags are the only ones worth standardizing and I actually like them. Maybe it's coincidence that these are the ones the 'analysts' don't understand.

The only point in standardizing the other diagrams is so that they could be loaded in to any modelling app. Most UML apps don't even agree on what is/isn't UML anyway so why have them?

Re:UML for HUGE projects.

[A55M0NKEY]

I agree UML is fine for HUGE projects as long as HUGE is defined as 'just a little HUGER than the one we are working on'.

24 years ago at Grumman Aerospace

[callipygian-showsyst]

I remember back about 24 years at Grumman Aerospace.

Back then, it wasn't UML, but "Data Flow" diagrams that were all the rage. We all took this "Yourdan" course and learned how to draw data flow.

What happened in practice was Management (including me ;-) drew these pretty little spider diagrams that were in a big thick book that pleased the Navy, and the programmers would write the actual working code that had little relation to the Yourdan diagrams.

My brief experience consulting with IT groups today that use UML reveals the same patterns. Some people earnestly draw UML in the beginning, and tools automatically make UML diagrams based on the code, but other than for top-level interfaces, nobody cares too much about them. It's a checkbox item.

(I'm glad I'm not an IT guy working in a UML sweatshop!)

I wrote a book on UML, but use it lightly

[MarkWatson]

Paul Harmon and I wrote a book on UML about 5 years ago.

At first, I thought that UML was a godsend because it did away with 12+ different modeling languages.

Still, for most of my work, schedules are very tight and my customers usually want to spend as little money on development as possible, so I find myself only using what I consider to be the highest value diagram types: use cases, very general class diagrams, and sequence diagrams.

-Mark

Re: Everything, including tools, in moderation!

[Anonymous Brave Guy]

IMHO, s/use/rely on/

I take your point, but there's nothing to say that you haven't just found a new context in which an existing pattern can usefully be applied. (After all, by definition something is a pattern in the first place only if it's got multiple real world applications.)

Hackles cartoon

[Ellen Spertus]

UML Gone Wrong

Don't go chasing waterfalls...

[Anonymous Brave Guy]

On the other hand I have seen environments where the UML was adopted as sort of cornerstone of top-down approach: all the design is done first and captured in UML.

That sounds suspiciously like the waterfall model.

The problem is that disagrams become absolete very quickly and people stop referring to them which accelerates the decay even more.

And that sounds suspiciously like the big problem with the waterfall model...

Why UML is bad...

[MattRog]

I'll save the typing and just link:
DbDebunk.com

Re: Everything, including tools, in moderation!

[Salamander]

When the design finally fails because you were missing something, the egotistical designer then blames the method.

...and when the method fails the snake-oil salesmen blame the implementation of that method. UML bigots are exactly like Extreme Programming bigots; they believe the tool is a panacea, so any failure when using it must be the result of not using it properly. It's a great way to count all the hits and ignore all the misses, and for people who couldn't make a buck by selling drawing programs as drawing programs to make a buck by selling them as something much grander.

Re:is there a free UML tool

[scharkalvin]

http://uml.sourceforge.net/index.php

uml2code...

[StarBar]

Unified Modelling Language is exacatly that, a unified way to build system descriptions with. Before UML people battled themselfs blue in the face about weather clowds or elipses were to be preferred when drawing systems graphically. The big problem with UML & co is how to generate code *and* get code changes back to the design at the UML level. No, you don't want to redraw the design and regenerate the whole program structure just because you found a small design flaw. Much of this requires expensive method specific tools (ie Rational's) that will wall you in like a whale.

What I would like to see instead is an Open Source command line tool uml2code that will generate source code or patches to an existing code base from an XML description. This would not work unless there is a reverse tool code2uml to analyze code changes, with necessary markup of cource, and generate a change to the XML description. There are some work done in the area but no tools that I know of. This aproach would give the benefits of a common XML interface between the GUI and the code generation and also the freedom to work with any normal set of compiler and tools on the output. Hmmm, just dreams though...

Here is a really easy methodology:

[master_p]

I have used UML in the past, but it is an overkill for most projects. Here is a quick altenative which works and can be applied mechanically:

1) write all functionality down as text
2) scan the text and:
a) note all nouns
b) note all verbs
3) make the nouns as objects
4) make the verbs as methods
5) write a dictionary.
6) discard what's left.
7) find the common parts between classes (find base classes)
8) draw an inheritance diagram.
9) draw an ownership diagram (the object model).
10) make up some examples and act them out (either by yourself or with a team) to see if anything else is needed.

It has worked for me quite well in the past for most cases. You can also repeat any number of steps any number of times desired, if cyclic developement is required.

Re:Here is a really easy methodology:

[Tablizer]

2) scan the text and: a) note all nouns b) note all verbs 3) make the nouns as objects 4) make the verbs as methods

This tends to assume that the relationship between nouns and verbs is one-to-one (every verb has a single owner noun). It is often not so in the real world. And the ties are often temporary. This often results in excess coupling. It hard-wires soft or temporal relationships into the code structure.

7) find the common parts between classes (find base classes) 8) draw an inheritance diagram.

Relationships between common parts often are not tree-shaped in the real world. Set theory would be more powerful for this than subtype theory, but nobody has figured out how to make polymophism work smoothly in set theory. Thus, one pretty much has to abandon OO inheritance and polymorphism to get the power of set thoery.

Re:is there a free UML tool

[Gramie2]

Poseidon is a free (beer) tool written in Java. At least, the community edition is free; you have to pay more for advanced features.

Re:Yup, it is better against earthquakes too.

[plastik55]

Rammed earth walls will survive much stronger earthquakes than frame construction

15,000 recently dead Iranians dispute that claim.

Granted, you can build earthworks that are stable with modern technology, using metal and concrete reinforcements and a better knowledge of materials science.

And one day, software engineers might be able to write a pile of C++ that doesn't fall down when pushed.

I've got UML Fever

[Jafa]

and the only prescription is more cowbell!

(not funny to non-snl fans).

J

Re: Everything, including tools, in moderation!

[eyeye]

What book is that?
Thanks.

Same problem exists for other technologies

[jfrain]

People can, and will, abuse just about anything.

Many of the "meta-fevers" described in the original article are not unique to UML. They are symptomatic of how people approach adopting many kinds of technology. I could easily make a few substitutions in the article, and poof, I'd have a "Death by XML Fever" article.

Design is good. UML is just a tool that can aid design. There's no "one size fits all" for how UML is used by people or in projects.

The amount of development process required what process is actually necessary, and how the process is implemented and enforced is unique to each organization. UML can be a part of that process, or not. Depends on the people in your organization and what seems to work best.

Ohboy! Now you can have pretty pictures to ...

[Ungrounded Lightning]

I'd like to see a true OOPs language like Smalltalk (OOPS are strong on prototyping) and UML (Strong on graphical modeling) combined with that tool mentioned awhile back that converts sketches to a refined diagram (Tablet PC's or Electronic Whiteboard). Throw in ad-hoc annotation capability and we are almost there.

Oh, boy. Now we can engage in rabid prototyping and have pretty pictures on the wall, too.

Bring up Smalltalk - or other rapid prototyping systems - and you demonstrate that the methodologies are leading you into the design-religion morass.

Rapid prototyping has been responsible for a plethora of project failures. This is because it consists of attempting to code without a design. That works if one person can keep the whole thing in his head, and breaks when the problem is large enough to need partitioning.

Rule of thumb: If rapid prototyping doesn't complete in 30 days, the project is to big for it. You must fall back on other design methodologies to succeed. Or you can continue for as much as a few years but you will eventually fail.

What you're talking about appears to be hanging automated tools on the rapid prototyping system to automatically infer the design document from the code. (Essentially, you're hanging the skin of a formal design methodology onto rapid prototyping in the hope of patching some of the latter's problems.) This is a cart-before-the-horse approach. But it also backs into the "self-documenting code" and "provable correctness" fallacies.

(It also reminds me - in style if not in substance - of a certain wild-eyed Smalltalk visionary's claim that software testing methodologies would have to improve until they could test his design in less than 30 minutes - because that's how often he changed it. B-) )

The key to robust code is to design it TWICE: Once as a human-readable spec, again as code. One must NEVER be automatically generated from the other - because debugging doesn't find bugs, only disagreements between specified and actual behavior. A particular implimentation of "cat" is VERY buggy if what you wanted was "ls". This is why correctness-proof tools aren't a panacea (and are misnamed): They work from a spec of what is correct, and writing a formal spec is ALSO a programming problem, subject to error.

It's OK for both the doc and the code to be in formal languages. It's OK for there to be a set of tools to perform part of the comparison between them automatically. But if you generate one from the other automatically, the only thing you can test is your generation tools. If it doesn't do what was intended that won't show, because the spec and the code will both prescribe the same "broken" behavior. And it's OK for a human to work from a spec toward code - because he'll be giving critical thought to the spec as he codes, and thus tends to discover spec errors.

But it's important for the two languages to be as different as practical, to put the designer and the programmer (especially if they're the same person) into different mind-sets when writing each. The more different they are, the more likely it is that problems will be discovered during the generation of one or the other. (This is why correctness-proof tools are useful despite being misnamed.)

In a real-world project the code and the design co-evolve to a significant extent. When you find a bug it's usually in the code (because the design went ahead of the code and had much standalone debugging). So you fix the code. But sometimes writing the code exposes a bug in the design. Then you fix the design. When you're done the code and design agree - and they're both near-perfect because the final versions of both withstood the entire trial-by-fire and many-eye inspection.

Why Modeling (and UML) could be useful...

[ndykman]

Well, I'm one of the authors of the UML 2.0 specification, so karma be damned. Here's a totally biased response.

1. UML enables communication on design, architecture that everybody can contribute too. The standard notation and concepts in UML 1.x enabled a foundation for communicating in a team about aspects of software that are not easily gleaned from the code.

I think this in itself makes standardization worth it. Also, it's not a secret language of architects. You read the same books they do.

2. You don't have to have all in one place. The nature of modeling allows models to be developed across different diagrams. UML 1.x is incomplete in this area, UML 2.0 does better. Pushing this notion could lead to some to some interesting places, one could be AOP without requiring AOP code.

3. Some changes are easier in UML. Tools don't do a good job of this, but I find it much easier to change object hierarchies and relationships in UML than in code. Imagine extending this to other areas besides the static layout of classes.

4. Metamodeling. The problem with UML and it's tools right now is the complete lack of metamodel extensibility mechanisms. It's like XML, but you can only use a fixed set of schemas.

This could be a real rat trap, but on the other hand, it could very cool. For example, imagine extending the base class metamodel to add what your project needs for persistence, integrity and object communication, and instead of writing code for every class to enable your special features, you use a model tool and templates to automate most of the process for every class.

5. Little Languages that everybody can use. If metamodel holds promise, it is basis for providing Domain-Specific Modeling Languages that take advantage of common metamodel concepts and visual syntax to reduce the learning and usage curve for every language. Having standards in this area help ensure interoperability and lack of lock in.

Oddly, Microsoft is the only vendor right nowthat really seems to be taking on the notion of metamodeling and DSMLs. I expect IBM and others in Eclipse to do the same around the EMF.

6. Modeling could be a complementary abstraction to programming languages. With some exceptions, we rely on code to produce systems. In some areas, models can often provide additional information that is not "immediately clear" in the code, and can automate the generation of that code.

An excellent example is E/R modeling. I would argue that E/R modeling serves as an good tool for designing relational databases, and shows things about the database that may not be clear from the set of DDL statements.

Now, imagine having the ability to create a whole set of these models that all carry a common infrastructure and tool set. Your DB modeling tools is your XML modeling tool is your OO modeling tool is your Workflow modeling tool, and so on.

The problem is that I view UML as "modeling middleware". I don't see it as just a notation, but I see it as a core infrastructure to base modeling tools on. This probably because I worked on the metamodel. In other words, I've spent too much time inside UML that I see the outside much differently.

Granted, most tools make it seem that UML is pretty (expensive too) pictures. But, hopefully, with UML 2.0, people will understand the real promise of UML and modeling. I think it goes beyond the surface syntax.

Now modeling is not UML. In fact, if Microsoft really pushes forward with Whitehorse, they may create the de-facto modeling standard.

The UML community becomes much more aggresive about providing metamodeling capabilities. Also, XMI needs to improve big time. Also, the OMG and UML would be well served by reaching out to MS and staying in tune with where they are headed, so they don't get caught totally off-guard. There is hope, I think MS has does some good work for the W3C (and some bad work, of course).