I thought perhaps we could start a thread just describing (in brief) what we expect to get from using the UML and as the thread and time progresses how our ideas/actual uses change.

This would give us, as developers, (considering to or) using the UML an insight into what others are doing and how we all, as a community, can develop from these insights?

Mark Saunders
 

i suppose it would be inpolite of me not to note what on earth i am doing/intend to do with the UML.

predominantly an interaction designer and secondarily a programmer/analysts i would state my main use of the UML is as a communication tool.

i hope to use the artefacts (probably use-case diagrams, interaction diagrams and state-charts) of the requirements elicitation and development activities as leading documents for the design of usable interfaces.

if the documentation created earlier in the development process can be used directly then this would

• save time in terms of developing the interfaces


• reduce any repetition of work


• ensure that the interface(s) allow the user(s) to complete the tasks indicated as necessary in the development process

i guess basically i want to see if we can bridge the gap between the software development and HCI practices and what extensions etc. would be necessary to do so.

cheers

Mark Saunders
 

>>
i hope to use the artefacts (probably use-case diagrams, interaction diagrams and state-charts) of the requirements elicitation and development activities as leading documents for the design of usable interfaces.
<<

That sounds reasonable.


> and what extensions etc. would be necessary to do so.

The extensions, if any, are largely dependent upon your problem domain. The reason UML is extensible is exactly that every possible need can not be accounted for.

-pete

I use UML artifacts to explain, graphically, to the business user and developer what the system needs to do. I find it easier to design the system using use case diagrams, activity diagrams before we go into coding.

I am finding that the majority of my Software Develoment Management process problems can be resolved by implementing some of the Rational Unified Processes, so I am a big advocate of them and of R'Rose.

The first time I used UML to describe to a business user how their needs were to be designed as a system they were totally floored and understood exactly (more or less :)) what development was going to do. Their modifications to the system, during design, was done by reviewing roles of actors and use cases. Also the time to solicit requirements, do the design and get into coding was more then halved. The quulity of the resultant code, was very good. All of the requred functionality had been met, testing was simplified since testing had a clear view of requirements. We hit our budget and timetable right on the money with a 30 man month project.

My 2cents.

Cheers,
Ivan

In not now, when?
If not here, where?
If not us, who?

Just do it!!

pivan, in you statement that ..The quulity of the resultant code, was very good. All of the requred functionality had been met.. this is one of the often advertised benefits of the RUP (which i can truly accept) but in terms of usability of the system - did you do any other testing?

the reason i ask is that the angle i am coming from is that i understand much of the software development process and resultant code to benefit from using the UML but what diagrams etc. will help us to measurably assess how usable the system is? - how far can current UML compliant models etc. go to confirming this with the user.

we can use the Use Cases as an, almost, contract between ourselves and the users - such as "..providing the ability to list all stock items of a particular category.." but how do we test that they (not the code) can do this efficienty. from general reading it seems that the dominant tool is the Task Analysis model - but using task analysis would appear to be a difficult/time consuming method. have we any advice for, perhaps, integrating such models into the UML and utilising work already done/to be done?

sorry for the somewhat verbose message
m

Mark Saunders
 

Gosh Mark... You raise some really intersting points.
First, I am not familiar with Task Analysis model...but I will be, and I thank you for this. I still have much to learn about UML.

Second, the could not the ultimate system test of the usablity of the final system be determined by benchmarking against other like systems? Or by how well it is accepted by the end user? Of course all of this is after the fact.

Is the Holy Grail how to do the above in UML?

I am interested to see what other responses this thread recieves.

By the way, I find this all to be of enormous value.

Cheers,
Ivan

In not now, when?
If not here, where?
If not us, who?

Just do it!!

> what diagrams etc. will help us to measurably assess how usable the system is?

None. I don't quite understand the thought behind the question.

'Usability' - The state or quality of being usable.

How can a 'modeling language' measure that? Can you provide a single definition for 'usable software' that can be used to assess every applications user interface? I don't think so.

Different application's target user base and environments are extremely varied. You can not use the same measure for an office word processor application that you can for a fast food restaurant cash register application.

"But, that's just my opinion... I could be wrong".
-pete

palbano - that's very true. so by definition one of the most important aspects of designing a software system is to know the user.

this is where we utilise the use case - performing the initial requirements elicitation.

where do we go from here? - i realise there are infinite options, what i thought might be useful was if people post the route they would choose and why.

for example - task models to agree the functionality with the user, sequence diagrams to confirm what objects/parts of the system do what and who they need to communicate with, state-charts etc.

in terms of agreeing usability what i wonder is this -
do many organisations agree with the client the functionilty of the software.
  if so how?
how many go one step further and agree that tasks will be performed in some form of measurement
  if so how are these measurements derived/agreed
what part do the recommended/addition models play

this form of measurement i guess we could refer to as the usability (i do, of course, realise that just because a clerk may input an order in 5 minutes doesnt necessarily deem the system to be 'usable') for the purposes of our discussions?

work, family, sleep - choose two!

Mark Saunders
 

In OOAD, how should the process of development be? Should use-case analysis go first then the identification of classes and objects?

Can anyone give me which process should be done first in OOAD?

Thanks!

~~~
"In C++, only friends can touch your private parts." - Emmanuel Ackouoy

The first thing you do in any analysis is to get the requirements of which use cases are the end product.

A good book that takes you through the whole process is
"Applying UML and PAtterns" by Craig Larman. Full of good stuff and lots of useful diagrams and explanations.

Thanks. After that then what? I'd like to know which steps to take first. Could you point me to a web resource of the "Applying UML and PAtterns", if there are any.

~~~
"In C++, only friends can touch your private parts." - Emmanuel Ackouoy

Do a search on google and you will find numerous websites.
OOAD is an iterative process split into various phases, each phase can be revisited in later development cycles as you refine your requirements, analysis and design.

But probably the most important part of OOAD is the creation of a Conceptual model.
Do a search for "Rational Rose" - you may find some good info there.