|
Moving an app from desktop to server, everything changes: We have different frameworks, different libraries, different tools, etc.
Even if you say I want to write an app that runs both standalone and on the network, where do you start?
Where is the mini browser that you can just point to a folder of php/asp/aspx files, without having to configure a port? (Silverlight makes the first attempts at running locally, niiice. Still, the straightforward experience is not yet there.)
.NET has tons of serialization code, database adapters, ORMLINKVOODOO, and whatnot. But where's the library that lets me write a local app, or I enter an URL/network path and BLAM! Everyone in the office sees and manipulates the same data. (SQL compact seems to compact closest to this. For which the IDE integration still needs a magic command line roundtrip to make it work).
Why, as a developer, is my first decision for a product whether it should be "web based" or a "rich client"?
Why as a buyer, is obe of my first decisions whether I want a "rich client and shuffle around documents", or "something thatruns on our local web server, and I need the nerd squad to install"?
I don't say it's impossible, but once you realized there are a lot of applications that you can try out and start with locally, and might to make available to everyone in the department later, it seems unecessarily hard.
|
|
|
|
|
All great points...
I may point you here[^] for a possible solution. The newer versions DO Offer an Out of the browser experience. I think it started slightly in V3 but has been 'fully' implemented in 4 and is going to be a focus going forward.
From here I think the lines between the online and offline experience [platforms is going to start getting blurry.
|
|
|
|
|
Ray Cassick wrote: The newer versions DO Offer an Out of the browser experience.
Interesting.
What started with AJAX and developed into SL, may now end up coming full circle back to a desktop app.
Now we will need tools to write more plumbing code for us, so when we open VS, and create a new desktop app, a blank form comes up that we drag controls to, double-click on them, and start writing code. I know, I know! Overly simplistic perhaps, but you get my point.
For example, something that would help us do asynch and parallel programming without needing a brain 3x the size. Anyone want a used brain? I'm looking to upgrade.
____________________________________________________________________________________
The Vulcan Science Directorate has determined that time travel is impossible.
|
|
|
|
|
Why not just make it a Click Once "Smart Client"? Then you can keep it a windows application, but deploy it from the web like a Java Applet. You'll probably need very little code changes, it's more of a deployment option really and simple to implement.
Unless you have a customer willing to finance the effort, it's probably not worth the massive amount of coding you'll need to do. You can do smart client in a day. The beauty of Smart Client apps is that you have a single install location (on the web server) to keep up to date. When the user clicks the icon, it automatically goes out to check for a new version and downloads it before starting up the app.
It's really best of both worlds...
|
|
|
|
|
Reminds me of something I did several years ago with a VB6 application...seperated it into logical modules, made each module a user control ocx, then hooked them up on a simple web page. The customer wanted a web application, and that's what they got...an spplication that is run in a web browser (IE only though) but behaves as a desktop application. In addition, the web page and controls could be run in offline mode (not so critical these days) and updates were automatically downloaded to the client. This setup worked extremely well for many years.
|
|
|
|
|
Thanks for this suggestion, kmoorevs!
Yeah, probably an easy to implement solution, but since this is a commercial app with lots of shiny UI, I doubt this would end up looking cool enough. I have no choice really - have to consider that aspect!
But sometimes I wish I'd kept the whole thing in VB6 anyway - trying to surf all the changes over the years has probably been the most time-consuming part of my work. And for what? I wanted to "keep up with the times". Or - maybe I bought into the hype. But I must admit that VB6 alone may have led to some dead-ends, so maybe I did do the right thing. Good ol' VB6, gotta love it !!!
____________________________________________________________________________________
The Vulcan Science Directorate has determined that time travel is impossible.
|
|
|
|
|
MachineGun wrote: Why not just make it a Click Once "Smart Client"?
Hmmm!
Do you have an article in mind that might help me analyze this option? Or would it be sufficient to wade through help do you think?
____________________________________________________________________________________
The Vulcan Science Directorate has determined that time travel is impossible.
|
|
|
|
|
This is a good overview of what a Smart Client is:
http://msdn.microsoft.com/en-us/library/ms998468.aspx[^]
This is a starter how-to:
Click Once Deployment Technique[^]
This article is missing one huge step though - Signing your app with a certificate and running the MAGE.EXE command on the deployment server. It'll work in VS, but then when you move it to a real web server, you have to include the certificate file and then run the MAGE commands to "certify" it on that machine.
For instance; Say your application is called "BobsApplication"
You would copy up the deployment files that VS outputs AND the certificate file (.pfx) that you need to generate under the "signing" tab. (in the Project->Properties->Signing tab), and the MAGE.EXE program up to your web server. You can put the cert. file and mage.exe in the same folder as your application. Then run the following MAGE commands at the DOS command line:
mage.exe -Update BobsApplication.application -pu http://YourServerName/BobsApp/BobsApplication.application
mage.exe -sign BobsApplication.application -cf BobsApplication_TemporaryKey.pfx
** Where BobsApplication_TemporaryKey.pfx is the certificate file that you generated under the Project->Properties->Signing tab. I also assumed you created an aliased folder in IIS call BobsApp in the first mage command. MAGE.EXE comes with Visual Studio, you can find it in the SDK folder (C:\Program Files\Microsoft Visual Studio 8\SDK\v2.0\Bin)
Then, to Run your application, you would go to the browser and type:
http://YourServerName/BobsApp/BobsApplication.application
I would also search Google for something like:
"Building your first ClickOnce application"
or
"Building your first SmartClient application"
Hope that helps!
|
|
|
|
|
Wow, thanks MachineGun, for going to all that effort - excellent reply!
I recently implemented an auto-update feature in my app, based on this article:
Application Auto Update Revisited[^]
The main app checks, downloads the update, and if that all goes well, launches another app. All the second app does is replace the main app's executable, and then launch the main app.
It seems to be working. I used that because I read some negative stuff about ClickOnce, I forget where. But I wonder whether you have a comment on whether the procedure you have outlined in your message has some advantages in your opinion.
Thanks again for the reply! I will analyze it! I noticed there was some command line stuff to do. I guess the VS installer project could implement that somehow. This is a commercial app, where I would prefer that the installer does everything so the IT folks don't need to do any tweaking. But that is only from a quick perusal of your post. As I say, I will take the time to read the references you included.
Bob
____________________________________________________________________________________
The Vulcan Science Directorate has determined that time travel is impossible.
|
|
|
|
|
Take a look at visualwebgui.
There's a screencast transforming a winformsapplication to a webapplication with their
softwarestack. 
|
|
|
|
|
Thanks Ralph.Popp!
I went right away to the site and visualwebgui does look quite painless. The price is right too.
Now this discussion was started to discuss strategies for making my app be a browser app, so please don't be annoyed if I include some philosophy here -
Maybe I'm wrong, but I'll bet their product requires me installing their dlls on my customer's server. Just so you see the workings of my tiny cynical mind: Since the visualwebgui folks are on version 6.4, that means they had versions previous to that, and in all likelihood, will also have versions 6.5, 7.1, and so on. My concern is that if I don't have the source code, I become dependent on other developers outside my "team of one" when there are problems with compatibility. Then my customer is mad at me, and then I have to get mad at the visualwebgui people, and of course then the problem escalates, and I end up sneaking around at night in my black ninja outfit, getting revenge by letting the air out of their tires. Ha ha. But seriously, I really try to do it all myself if possible. That philosphy has served me well, although at the cost of my time, a lot of it, which I realize could be saved through using a more "convenient" 3rd party app. In fact the same goes for those devexpress free controls. I can't use them, cool as they are, because I have no control over the source code.
Let me know your thought on this, if you see a gap in my reasoning, and thanks again for your reply.
____________________________________________________________________________________
The Vulcan Science Directorate has determined that time travel is impossible.
|
|
|
|
|
I would say before you dive in, restructure your code base to enforce a presentation layer, and if feasible (or not already accomplished) pull all the presentation layer off into a separate solution. this will make the development of even code behind faster. if you can, get everything scrunched down to one form, with all the necessary controls displayed on that single form.
once you've accomplished this the mapping of winforms to webpages is quite simple (and if your comfortable spending about a week slopping code around flagrantly) you can etch out a crude close approximation first gen WebApp rather quickly. Really the "functional template" has already been written (legacy code) and you want to float a new UI on it, so Hack together a minimal UI (in appearance only) and then focus on getting everything to stitch together...
I'd blame it on the Brain farts.. But let's be honest, it really is more like a Methane factory between my ears some days then it is anything else...
|
|
|
|
|
ely_bob wrote: restructure your code base to enforce a presentation layer
My wife and I were discussing this last night, and that is exactly what I realized would be the first step!
A windows form has 3 files, the resx (which is the pics, which a web page also fetches), the designer, which is the layout, which a web page also must generate, and a code file, which a web page also has.
If I create that 4th file, a class that has all the real code in it, that is the file that will ultimately go on the server side.
ely_bob wrote: pull all the presentation layer off into a separate solution
Hadn't thought of that point of putting it in a different solution. Perhaps you could share the motive for separating it to that degree.
Excellent practical advice, thank you!
ely_bob wrote: all the necessary controls displayed on that single form
Why a single form? I'm sure you are saying this from experience.
____________________________________________________________________________________
The Vulcan Science Directorate has determined that time travel is impossible.
|
|
|
|
|
BobishKindaGuy wrote: Why a single form?
Well this makes setting up the PageBase class easier and allows for placement of session variables(or however you decide to persist user info).. all this will happen as a approximatively 1-1 mapping from a single form.. (basically this will allow you to write less code in your webapp.
The reason for a separate ui solution(s) is common in the XNA community, it allows/forces all the (business) logic to be centrally located, and is just a safeguard to taking the easy road as opposed to conforming to a set of best practices. this will also allow you to do a compare against your existing code base, as long as you follow the basic solution and page layout (can be very handy for debugging unexpected behavior).
I'd blame it on the Brain farts.. But let's be honest, it really is more like a Methane factory between my ears some days then it is anything else...
|
|
|
|
|
Thanks again, ely_bob!
I have quite a few tabbed-dialogs and wizards.
What do you think of the idea of a webpage that aligns those vertically so the user work their way from the top down?
There's a notepad replacement app that does that, and it looks really cool. (Of course, that app is a desktop app ha ha ha)
Regarding the separate solution for the UI, do you mean separate project or separate solution completely?
Either way, I guess the code in the form would look something like (air-coding here):
Public Class WizForm1
Private m_UI_Handler As SomeUIHandlingProject
Private Sub Done_Click (e,sender) Handles btnDone.Click
m_UI_Handler.WizForm1_Handlers.DoneClick
End Sub
End Class
____________________________________________________________________________________
The Vulcan Science Directorate has determined that time travel is impossible.
modified on Thursday, July 22, 2010 5:35 PM
|
|
|
|
|
well Personally I hate to do more vertical scrolling then absolutely necessary...
I'd recommend either a presenter approach (the page reloads with what you want to present) or use a layout similar to what you describe, and use tags to hide/show just "like" you would be doing in a tabbed document(this makes layout easier and I prefer the approach because it is a lot more straightforward.)
If the wizard is a fundamental part, you may want to treat that as a flash app(however I am not familiar with these approaches), that way you can take control of the entire screen, grey out the background browser, and have a specialized solution just for your wizard. and this would be plug-able into your webapp...so...
I meant a one solution with ~3 projectss, a library(or folder of libraries), and 2 UI: winforms & webapp. then if you want to make some kinda mobile app down the road you can just add it to the solution, and add some dependencies and probably reuse a lot more code that way...
I'd blame it on the Brain farts.. But let's be honest, it really is more like a Methane factory between my ears some days then it is anything else...
|
|
|
|
|
Cool!
Your suggestions are definitely part of the solution to this problem. Isolating the UI from the rest of the logic, still within the desktop app, will make it a lot easier to move everything to a web app. The hurdles I encounter during this process will probably force me to think it through with the web app in mind. (E.g. passing certain parameters along with the call, etc.)
Thanks for taking the time to make these suggestions.
____________________________________________________________________________________
The Vulcan Science Directorate has determined that time travel is impossible.
|
|
|
|
|
I've been intrigued by Morfik. I was "forced" to learn TurboPascal to support manufacturing systems and that migrated to Delphi which I use for my own personal work apps (nothing commercial). The ability to reuse a large part of the existing code is very attractive to me but I haven't really studied Morfik in depth to see how easy/hard a conversion could be.
Just a thought...
|
|
|
|
|
|
Machaira wrote: Why does everyone think creating a web app is the solution to every problem?
Good question! I wish I didn't have to consider it.
But there are some good reasons. See my answer to Ray Cossack.
____________________________________________________________________________________
The Vulcan Science Directorate has determined that time travel is impossible.
|
|
|
|
|
Thin client apps are so much easier to administer.
1) Changes to the app don't have to be rolled out to 100's of possible users by installing something on their computer. Then when 98% of them are updated and you start getting calls from people who don't read email or don't click yes to update their software when prompted you start pulling your hair out.
2) No installation involved so it involves less helpdesk time.
3) Less worry that someone may leave your company with proprietary software on their laptop.
4) The ability to log in any where any time from any computer.
5) Less of a requirement for users to have high end computers because the load is transferred to a server or a cluster that is centrally managed.
6) the list goes on...
I speak from experience having to currently support and develop desktop and web applications. If given the choice, 100% of the time I would prefer to make a web application from a support standpoint. The only time I would consider a desktop application is if it were something that was not used by many people, was calculation intensive and there were no spare servers available for it.
|
|
|
|
|
Well put, MatrixDud !
Now if only what you said had been part of a course in high school, my app would have been in the cloud long ago!
But I went down another road and now, I've got this huge pile of (cool) code to push around into other smaller piles, I guess. 
Thanks for taking the time to share your very relevant perspective to this discussion!
Bob
____________________________________________________________________________________
The Vulcan Science Directorate has determined that time travel is impossible.
|
|
|
|
|
Correct me if I am wrong, but it seems to me that web applications are mainly for publishing type of applications. In other words, if your desktop applications involve a lot of user interactions, then they may not be suitable for the web. Otherwise, it is not too bad to port them using either of the current available technologies.
|
|
|
|
|
Cloud computing or SAS proponents would say the opposite. Salesforce.com would be considered very user interactive.
|
|
|
|
|
George from Saanichton wrote: web applications are mainly for publishing type of applications
My feel for this is that, like it or not, things are generally moving to the cloud (the web). Whether that ends up benefiting humanity is another question. I'm quite cynical about the whole thing. But it's not so much what kind of app is suitable for moving to the cloud. In my mind it's about the IT people in an org. They want to adopt applications for their org that can be supported centrally, not on every user's desktop. So the browser is the ideal desktop app - it is conceptually a window into the cloud, where the real app resides, and needs no updating. When there's a version update, the app vendor (me) just sends them an update to the app on one of their servers. The user receives the benefit of that the next time they launch the app using their browser. There are other benefits (and drawbacks) too, and CodeProject and many other sites are full of those discussions I'm sure.
Thanks for taking the time to share your thoughts on this!
____________________________________________________________________________________
The Vulcan Science Directorate has determined that time travel is impossible.
|
|
|
|
|
General 
News 
Suggestion 
Question 
Bug 
Answer 
Joke 
Praise 
Rant 
Admin 
Submit an article or tip
