Michal Harezlak

Ranch Hand
+ Follow
since Jul 06, 2000
Cows and Likes
Total received
In last 30 days
Total given
Total received
Received in last 30 days
Total given
Given in last 30 days
Forums and Threads
Scavenger Hunt
expand Ranch Hand Scavenger Hunt
expand Greenhorn Scavenger Hunt

Recent posts by Michal Harezlak

Originally posted by Nakata kokuyo:

- forgot do validation (for date, number field)
- testing not good enough, and user easily get error when use the application
- application not match to domain logic
- and so on for coding(web)

is that anyway to improve these ? thank you for guidance

If I could make my self immune to these problems I would probably be a famous person already.

However it seems that TDD, Unit testing and design by contract could be of help to you in regard to concerns 1 and 2. When it comes to 3. It seems that you could try to develop and manage your requirements or/and use XP's "on site customer".

For Design by contract see:

For Unit testing.

For XP:

MVC is a pattern (best practice) so I would get you familiar with it to get best possible understanding of it. Then you go and choose a framework or write your own. Then you write guidelines what to be aware of when using your framework and make it known (difficult part). In case [e.g.] you choose Struts, it could look like this:
Things to remember when with Struts:
�Only one instance of Action class is created
�Watch for thread safety
�Use local variables
�Do not synchronize your method if you do not have to it can be major performance bottleneck.
Hope that helps.

Originally posted by Vinod Chamoli:
What is the difference between Framework & Architecture.

It is rather a question of semantics. IMHO: Architecture is a set of frameworks and the interfaces between them. They are really very similar conceptually but architecture contains frameworks. You can speak of architecture of a framework. In very remote way it is like building architecture that encompasses all the sub systems (electrical, air, plumbing, telephone, TV and such) of the building.
Hope that helps.
Declarative security is great in EJB, no question about it. There is a slight problem that seems to be explained in literature with money related examples. Class of the problem: How to implement security on task that requires different security profile for operation involving different amount of moneys. For example junior trader is only allow do to orders up to 100K. In this case it is advise to do it programmatically and entangle business code with security. How about writing two separate methods, one for orders up to certain limit and the other for above. I am not suggesting to hardcode 100K in the name and functionality of the method. I am suggesting that leaving check for the limit inside the method and separating security aspect with declarative approach could be a pattern to follow in this example. Any other examples where programmatic security is necessary? What do you think?
Best regards.
All very valid points, thanks for your contribution. Lets start from the easiest one. I am against printing anything. Since I cannot afford RequisitePro or any other such expansive (license + training) tool. I have manufacture small document management engine (version control+web publishing) that is how I intend to avoid having printed out of sync documents. My personal experience is that �dullness� of document can be avoided by applying author�s commitment (read the thing at least once and remove all the crap that has no meaning) as well as validation (do we need it?) and verification (is it what we said it should be and dose it do what we want it to do?).
Yes I think I will need to document performance, acceptance, usability, configuration, documentation testing approachs. I see it more as a giving direction to QA and taking as much as possible of my developer�s shoulders (an example would be giving QA unit test data to create and analyze metrics). Showing them tools that could be used and explaining wher quality fits into development architecture.
I think that improving the process by analyzing metrics is possible. Lets say that metrics shows that even thought we have height number of unit test we still have high number of defects (of sort that unit test should discover) while doing test in labs. Lets go back and see how developers write unit tests. Perhaps we will discover that they do not have proper data set to use for testing. We could get this data from QA. Voila we have process improvment.
Best reagrds!
[ March 08, 2004: Message edited by: Michal Harezlak ]

Originally posted by Ilja Preuss:

Well, I am also not sure what you are getting at, but here seems to be a misunderstanding. JUnit is *mainly* a tool for the developers, not for the QA guys.
It's still a good idea to involve them early, though. In Extreme Programming, for example, you would want to have them help the Customer come up with Acceptance Tests using tools such as FitNesse.

Thanks for you comments. JUnit is mainly, perhaps only developer's tool. Although I see software development as a never-ending stream of development cycles (never mind the specifics of process). Developers work is never done. Unit test are alive in QA environment and they could be run there. Therefore an explanation of unit test could be beneficial for a QA person. They understand what it is and who to ask when something happens with it. Perhaps it is good to include them in Cruse control mailing list of daily reports so they can use it for metrics. Dose it make any sense to put this perspective inside SAD?
Best regards.
[ March 08, 2004: Message edited by: Michal Harezlak ]

Originally posted by David Peterson:
I'm not 100% sure about your question. [...]?

That is actually also my problem. I have this nasty habit of saying my ideas out loud to see how stupid they sound and what exactly do I mean. Thought and ideas have this amazing quality that they look excellent before being said.
With that said. I ma convinced that Test plan is necessary but it is not a part of SAD (Test plan as an instance of a test process definition). I have a feeling that there is an aspect of it concerning building blocks of environment and solution itself. JUnit as a technology and what it means for a QA guy (have them involved since start), Log4J, JMeter, cruisecontrol, ANT and such.
So I am not really asking about Test plan, rather about Architectural QA View of the solution and what should go in it and if that makes any sense?
Thanks David!
Unfortunately things too often are not what we like it to be or what we call it which is a root of all evil AKA miscommunication. I am considering adding a Quality Assurance view to my SAD (Software Arch. Document). I do not think that the name is as important (or perhaps there is a better name for it and I would definitely like to hear it) so the name is not as important as intentions. I see View idea as a grant element of any SAD because of its targeted nature. Creating suitable view by putting proper information in it is a key to communication, which is a corner stone of any human effort (see Tower of Babel biblical example). Well, My QA view idea is sort of vague at this point and I know it has to do with setting a common ground for QA and DEV teams and telling everyone what tools/software components we should use and what is this TDD and performance and stress and all the rest. I would like to hear your opinion on this idea.
Keeping the server stateless. Stateless servers scale well, as a matter of fact they scale excellent. There is a slight problem with it though. The session is kept on web tire (web container) or some other close to client tire. You have to redo the code for each client. You have to break business (aka problem domain) and put parts all over the deployment components.
Using stateless session beans and JMS with MDB seems to be very scalable approach but it seems that we are actually going back to transaction monitors like Encina or Tuxedo.
Any good paper on scalability that you would recommend?
MIDP vs. WML. I had this dilemma on my last project it went down to MIDP not being easily deployable and maintainable on all kinds of phones (it could be anything but my app was for phones). WML is totally client independent but lacks well defined interface. Exactly same problem with swing and HTML. MIDP is of course the choice for all off line capabilities.
Ok. Lets get back to my problem! I cannot put my finger on it put there is some problem with Swing Web based clients, perhaps it is this proprietary protocol that needs to be developed for communication with the server. BTW I remember seeing something in the vision document [or somewhere else] about enabling my users doing work off line. Is it really necessary and good idea to spend money all this hassle with keeping data up in sync. IMHO if you are at a place that dose not have net hookup you should not be working anyway.
Best and thanks again for all your help.
BTW I came across this little thing called WebCream http://www.creamtec.com/webcream [no it is not a porn outfit] it lets you port Swing apps interface to web with no effort and there is something called XUL (XML UI Language). I must investigate.
[ February 21, 2004: Message edited by: Michal Harezlak ]
Specially thanks to Frank for Mart Tech link, must check it out ASAP.
I see three aspects to this decision.
I will need to have HTML base interface no matter what. This app has front and back office. Front office is traffic intensive but with simple interface. So using load balancer and web container + sending messages to (MDB) search and submitting loads of data should work like a charm on front office side.
While back office is transactions, processes and basically used by a user throughout his entire day of work.
Deployment seems o be not a problem when employing WebStart although it adds to complexity and managing complexity, specialty the more complex ones is not easy.
Keeping it simple
Introducing two different sets of tools for developing UI is not wise. Needs more resources and discipline and in result is costly.
Productivity :
I need productivity TOOL that kicks [Three character word]. I can much easier get it done with Swing applet then designing HTML with GUI artists and such. Their designs needs to be redone with JSP/Struts/Sofia guys. It takes time and is tricky but looks prettier and my users are used to HTML interface, I still suspect they would love Swing, this must be validated with prototype.
I cannot make up my mind!
Thanks again!
Thanks a lot for your valid suggestions. It seems that WebStart mentioned by Thomas takes care of problems with deployment aspect of maintainability (run time environment and versions) therefore TCO should not be affected here. I intentionally used �fatter� instead of fat client. That might be a crazy idea, but I intend [not sure exactly how yet] to use it as a stateless client just for sake of benefits [is that proper English sentence ] that Swing gives in GUI development over HTML derivatives.
Thanks again.

Originally posted by Jim Henessey:
[...] Does anyone here have recommendations for a professional caliber UML editor that they've worked with which is either inexpensive or free/open source?
Much appreciated.

I was trying to get a discussion on that issue at IDE forum:
Unfortunately, no luck. Perhaps my thoughts will be helpful for you.
Best regards.
Ok, I am not entirely sure if that is a right forum to post my question. It has many aspects (I think), it is rather a question of architectonic nature. Please correct me if I am mistaken posting it here.
I am trying to decide on architecture of a solution that has rather complex UI in the back office component. This app will be hosted in one place and accessible by clients from all over the world. It will have to be accessible trough firewalls with not or minimal configuration at clients firewall settings (port 80 using https). I am trying to figure out if it is feasible to use applet swing client. I have not seen many projects using this approach as a mater of fact just one. It was rather complex time tracking/accounting system. I was not very impressing with speed of the interface at times (dual processor). I could not see any prohibiting risk when I thought about this yesterday. I am starting to suspect that this whole HTML interface (APS/JSP) thing is web designers conspiracy .
-Java Swing performance (I am not sure it that is still the case),
-Maintainability (client code will be on client workstation how to manage updates and version control WebStart?),
-Proprietary protocol over HTTPS and in case of use of XML considerable stress on client�s workstation resources ergo performance drawback,
-Sending just data not interfaces over the network therefore lower traffic (not sure if that is very important when using proxies on client side),
-Event driven interface with well defined and proven components therefore more intuitive, consistent and maintainable,
-Productivity tools available,
Any thought ideas, pointing out my faulty logic or misinformation will be greatly appreciate!

Originally posted by feroz muhamed:
from where can i get more information on value objects and data objects
[ January 28, 2004: Message edited by: feroz muhamed ]

Could be the same thing. Objects that contain just data have no behavior. You might also be referring to J2EE design patterns: DAO and Value List Handlers.