Re: Uses for D2JC (was: Gianduia and WO)
Re: Uses for D2JC (was: Gianduia and WO)
- Subject: Re: Uses for D2JC (was: Gianduia and WO)
- From: John Huss <email@hidden>
- Date: Sat, 7 Nov 2009 09:48:18 -0600
You could argue that it doesn't matter much if a JavaClient looks awful
on the Mac since 90% of your user will be running Windows - and
probably more for a business application. Also, the
Nimbus
look and feel introduced in Java 1.6 is pretty nice, and available for
Java 1.5 with a separate library. It is very aqua-like, but not
native, so you don't have to worry about it not quite matching the
native look, sort of like the way browser apps do.
Still, I agree that I prefer SWT's appearance to Swing's. You can also make a JavaClient app with SWT (not a D2JC one though).
John
On Fri, Nov 6, 2009 at 11:42 AM, Mike Schrag
<email@hidden> wrote:
Sure, I agree yours is real and mine's a picture, but that wasn't really my point. My point was about visual mismatch with the native platform. If you're OK with the app looking .... like Java ... , by all means, ship it. I find Java desktop apps to be mostly visually repugnant (the same reason that I work on Maclipse, but at least SWT has a fighting chance), and I don't believe you can actually make it un-gross without a huge amount of work. It's a question of what matters to you, though.
That said, pretty much everything in the main chunk of your email isn't really D2JC, it's just D2*. I'm sold already on the RAD value of D2* technologies, I'm just not at all convinced that D2JC is the right D2 to choose in most cases. If time-to-market is the primary concern, it seems to me that you have dramatically faster styling and design capabilities with D2W. You can hire an designer off the street to do CSS/HTML work for you, whereas you'd dump a pile of hours into custom Swing controls. This is essentially the same argument I have against the IB/Cocoa style of development. The RAD time in IB/Cocoa is just way longer than making the equivalent UI web-based. Maybe one screen you can layout a little quicker, but start doing multiscreen workflows, and you're just sinking time into Cocoa work (you can really see this when you try to RAD an iphone app -- i find it to be very expensive to do in contrast to web RADing). The flip-side, of course, is that it looks awesome. It's the trade-off again.
So I guess the question is really "given a project spec, what feature in the project would cause on to choose D2JC over D2W."
ms
On Nov 6, 2009, at 11:48 AM, David Avendasora wrote:
On Nov 6, 2009, at 9:54 AM, Mike Schrag wrote:
couple mins of Interface Builder mockup (mac-up?):
<PastedGraphic-2.png><PastedGraphic-3.png>
Nice looking and it shows how powerful Interface Builder can be in making UIs, but mine isn't just a mock-up it's a fully functional application - warts and all. It really talks to the server, sync ECs, validates, etc. etc. Again, I dropped an EOModel into my template project and launched it. Then used the EOAssistant to configure which attributes showed, what type of widget they used (PopUp, text field, explorer, etc) and where they appeared on the screen. That's it. No code. There's not a single Java Class besides Application, DirectAction, Session and two components: Main and JavaClient, none of which are modified beyond the stock template.
So D2JC is certainly _passable_, but it's far from being a proper mac app (those default old style boxes are the dead giveaway for java apps, though there's a Swing client property to switch box styles, IIRC). Now, certainly, if you spent enough time, you could tweak things and make it a lot closer, but it's always going to be off.
Yes, and so will any web app, only more so. Now, your point about the uncanny valley is very true. D2JC comes almost too close to the look of a native app. If it were further off like a web app people's expectations would be lowered and no one would complain. But being uncanny hardly seems a valid reason to throw away something that saves literally _hundreds_ of hours of development time, especially for internal, CRUD-type admin applications.
But I don't think that using D2JC to create shipping CRUD-type applications is really its most appropriate or useful application. To me, it's all about lowering project risk by getting stake-holders involved as soon as possible in the development process and as a prototyping and model-validating tool.
D2JC is an incredibly powerful tool when used as a means to an end. WIth D2JC you can have a _fully functional_ application with just an EOModel. This allows you, as a developer, to get a true feel for how well your model represents the business context without having to write any code. You can immediately start creating records in the database, setting relationships, deleting objects, orphaning them, etc. through EOF which will verify that the relationships are properly setup including that your deletion rules and ownership settings work as you expect. Again, you can do all this without writing any code! You don't have to ever show it to an end-user if you don't want to.
Change your model? No problem. Rebuild (copy the revised EOModel to the build directory) and relaunch your app and your changes are immediately reflected in the UI. Since you haven't written any code, you don't have to refactor! What's better than Eclipse's refactoring tools? Not needing to refactor at all. D2JC greatly speeds the process of modeling and validating that the model reflects reality.
On top of all this, since you have a fully functional application, with all CRUD functions, revert functionality, validation, etc, it _can_ transition into the initial data-entry, data-cleanup application to populate/scrub the database and even be used by non-developers (business users). How many times have you created spreadsheets for users to do data entry into so you can then import them into the DB? With D2JC what you give them can be a real application that writes directly to the DB.
Keep in mind that you can do all of this - requirements to working prototype - in ONE DAY with no code. It is easy to start with some basic requirements, create a model, then an app, and take it to the stake-holders to show how various pieces fit together and get even better feedback on workflows because you are giving them something they can see and understand not an abstract diagram. You can then make revisions and return with a revised, working application _minutes_ later. Yes, I said minutes!
Once you've validated that the Model accurately reflects reality by having actually putting it to use, you can now start the labor-intensive process of creating the finely tuned UI that looks and works exactly how you want it to. Because you have already done a substantial amount of model and business context validation there will be far less risk of having to tear it all apart due to having missed something fundamental in the model.
I'd like to see any other technology do that, WO-based or not.
And we're just talking chrome, not behavior of the app, which is often even more off.
D2JC certainly has some UI clumsiness to it, especially when working with many-to-many relationships and dealing with Inheritance.
Of course you have to do all these tweaks for each platform, too -- though you're a fair bit closer than my CocoaClient app is running on Windows :)
I think what it comes down to is that I would not suggest using D2JC in a situation where you have to have complete control over the UI or have even reasonably complicated work-flows. On top of that, I don't think it would be worth the effort to tweak the UI for one platform, let alone multiple ones. The fact that it runs "passibly" on all Java platforms without writing any UI code at all, to me is simply amazing and incredibly powerful.
Whew, that's more than I intended to write. :-)
Dave
ms
On Nov 6, 2009, at 8:50 AM, David Avendasora wrote:
On Nov 6, 2009, at 8:17 AM, Mike Schrag wrote:
So if we're stirring the pot here ... For me, it's not being cool in the browser, it's about Java apps looking terrible. You can immediately tell when you're using a Java app ... Things are just never quite right, but they try to sell themselves as being native. It's an uncanny valley situation. Eclipse/SWT are using native controls for lots of things and they get it wrong, too. Swing just doesn't have a chance. Go try IDEA -- it looks TERRIBLE. Look at their preference panels. I tweeted when IDEA went free that I can see how to make Eclipse right, but I'd never be able to make IDEA right.
I understand the uncanny valley. In most cases you are absolutely right. Here's a screen grab from my current D2JC project, I think it looks pretty good, although it doesn't look like a modern OS X app (iTunes, etc):
<PastedGraphic-3.png>
For browser apps, it's obvious they're not native apps, and the bar is set low in the browser at this point, so you can make a slick looking app, and it doesn't have to be perfectly native, and people are still OK with it. I would be far more interested in CocoaClient where you actually have a chance of pulling off a nice end-user deliverable.
I agree. For me a WO Cocoa Client is the Holy Grail of Client-Server. I want it. Badly.
That said, I recognize that there are plenty of apps where "looking slick" doesn't really matter -- that you just need to get some business app out there. But what does Java bring to the table that you're not getting in the browser?
With plain WebObjects JavaClient, you get
1) Real Java on the client. No messing around learning _javascript_ (or waiting for the "next great JS framework") to implement UI logic.
2) EOF on the Client with
- automatic syncing of Client and Server EditingContexts (works very similar to Child ECs)
- batching
- faulting
- validation
Why not just use D2W? Drag and drop is about the only thing, and that will be in the good browsers pretty soon.
For me, it's not about the features you get in the client, although there's some cool stuff that way too. It's the dead-simple development side that makes it so awesome to me. With WebObjects D2JC you don't have to write _any_ code at all. No HTML. No CSS. No _javascript_. Not even rules! You hear that D2W guys? The D2JC EOAssistant works great and it will write most of the rules you need! Here's a screen capture of it (it's also a Java Client app that communicates with WOLips to update the user.d2wmodel file):
<PastedGraphic-6.tiff>
Anyway, I didn't mean to hijack this thread with Java Client, but hey Anjo brought it up! :-P
Dave
I do, however, think D2JC is a _neat_ technology, in that it's amazingly clever what it's doing under the covers, I just am not sold on the end result.
ms
On Nov 6, 2009, at 7:41 AM, David Avendasora wrote:
Hey! Did I hear "JavaClient"?! :-D
Yeah, it will be cool if someday we get the tools to do something client-server like JavaClient.
Oh wait! We already _have_ WebObjects-based JavaClient, and Direct-To-JavaClient and it works today and has for _years_.
Sure, it's not as "cool" as a browser-based solution, in the same way WO isn't as "cool" as Ruby.
**ducks, runs for cover and scrambles to get the D2JC project template updated**
Dave
On Nov 6, 2009, at 7:32 AM, Anjo Krank wrote:
Not really sure... basically you'd need something totally new, as this would be more like JavaClient and not like a web app. But all this is *moot* until we don't have it.
Cheers, Anjo
Am 06.11.2009 um 13:05 schrieb Mike Schrag:
Yeah .. I suspect there could be a GianduiaLook, and that would make a lot of sesnse.
ms
I think we've seen with the SproutCore stuff that it's not. Apart from *maybe* a JS D2W.
Cheers, Anjo
Am 06.11.2009 um 02:54 schrieb Mike Schrag:
rom my perspective, I don't know that it's worth building a server stack on top of it as much as I think it would be nice to leverage the development tools we already have to make it easier to write the _javascript_.
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Webobjects-dev mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Webobjects-dev mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden
David Avendasora
Senior Software Engineer
K12, Inc.
*****
WebObjects Documentation Wiki : http://wiki.objectstyle.org/confluence/display/WO/
*****
WebObjects API: http://developer.apple.com/legacy/mac/library/documentation/MacOSXServer/Reference/WO54_Reference/index.html
*****
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Webobjects-dev mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden
David Avendasora
Senior Software Engineer
K12, Inc.
*****
WebObjects Documentation Wiki : http://wiki.objectstyle.org/confluence/display/WO/
*****
WebObjects API: http://developer.apple.com/legacy/mac/library/documentation/MacOSXServer/Reference/WO54_Reference/index.html
*****
David Avendasora
Senior Software Engineer
K12, Inc.
*****
WebObjects Documentation Wiki : http://wiki.objectstyle.org/confluence/display/WO/
*****
WebObjects API: http://developer.apple.com/legacy/mac/library/documentation/MacOSXServer/Reference/WO54_Reference/index.html
*****
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Webobjects-dev mailing list (
email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to
email@hidden
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Webobjects-dev mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden