Slashdot Mirror
Rapid Open Source Development for the Unix Console?
Adam Weiss asks: "With Web Interfaces and GUIs being all the rage these days, it seems hard to find much about console application development. Web Interfaces and GUIs look pretty and impress people, but I've noticed that it's awfully hard to beat the speed of a well trained operator on a well designed console interface. Some of the HR folks at work use a console app to access employee records while others use a Windows GUI. The console folk can lookup and update three records in the time it takes the the GUI folk to clicky-clicky through one. So, are there any mature Open Source toolkits that would enable rapid development of console applications. Sure, there's curses, but that's low level pain in the arse. I'm talking like something that is specifically designed for building database applications- kinda like an extensible version of Microsoft Access Forms for the green screen. Something that's pretty easy for the simple stuff, but lets you break out and get complicated if you need to. (unlike Access) I know there's gotta be plenty of obsolete commercial stuff that makes these kinds of projects easy. I just want to know if there are any Open Source alternatives that are somewhat modern and well maintained."
Links is a text-based web browser - building a console interface can be very simple using HTML and Links - you don't have to play with ncurses' mess.
The problem with GUI speed is not so much caused by design - it's the darn mouse that slows down everything... instead of doing a simple Ctrl-S to save, it's all these 6 steps (or more!):
Web Interfaces and GUIs look pretty and impress people, but I've noticed that it's awfully hard to beat the speed of a well trained operator on a well designed console interface
The key phrase here is a well trained operator. GUIs are pretty, and can be slow, but almost anyone can plop their butts down and start working on it immediately. Console apps, fast as they are, require sometimes days of training before working on it, then weeks (or months) of experience before getting truely fast at it.
This tradeoff is what businesses look for, and if its a spot with a high turnover ratio, they don't want to throw money away on training. Sometimes its just better to make it more user friendly than speedy.
I just wanted that point to come across for why companies go for web design. Right now I'm putting a J2EE front-end on a mainframe backend for a company to trade a 2 week training class to a 1 day relaxed training class to operate the software. It is something where speed is desirable, but web speed isn't that slow for the internal app.
Good quote, too many chars. Seriously, the slashdot 120 char limit sucks!
If the same amount of time and effort was spent making the gui ergonomic and "keyboard accessible", I suspect that you would find that their speeds were about equivalent.
I think GUIs take a lot more effort and time to do well than most people will admit. This is one reason the UNIX CLI is still around (it's quick, it's dirty, but it works reasonably well).
Mainstream GUIs often have these problems: the fonts tend to be small making them hard on the eyes, grey and white backgrounds with black text are hard on the eyes, there tends to be more information packed into one screen often requiring scrolling (time consuming and hard on the eyes), GUI layout managers are mastered by few leaving some pretty darn ulgy forms (again, hard on the eyes), many GUIs imply the mouse for intuitive navigation, etc.
It also doesn't help that GUIs are enormous programs, which are inherently unreliable and often not predictable. One reason Windows gets such a bad rap is that Microsoft produced a huge program in about 1% of the time it should have taken if they wanted to do it well.
Healthcare article at Kuro5hin
A well-designed GUI app is not necessarily less efficient to use than a well-designed console app. The problem is that many GUI apps, especially custom-written apps or vertical apps, are based on the horrible "forms" or "screens" paradigm, where the app is just a series of screens, each with lots of fields. (And all web-based apps are like this so I also classify them as "horrible", despite (or maybe because) the fact that I currently write web-based apps for a living (because that's what people like to pay for).)
In fact, given talented UI (GUI or console) designers and application developers, I'd guess that a GUI app could be more efficient to use than a console app at least 75% of the time.
And of course, a well-designed GUI app should be easier to learn than a console app almost 100% of the time.
This is actually a pretty good Ask Slashdot, for a change.
I know exactly what you want -- something like a text-based Hypercard...a front-end building toolkit. Also, the insight that computer operators tend to be significantly faster with text-based UIs is interesting.
Suggestions have included text-based web browsers (not really ideal...I think what the guy's thinking of is a screen-by-screen interface where there's no scrolling or anything, much like those custom DOS apps that banks use) and perl (AFAIK, perl will let you enter lines but doesn't natively have a great text-based UI tookit). I don't think any of these are really appropriate.
You may want to look at dialog. It's a GPLed higher-level toolkit that sits on ncurses that ships with (at least) Red Hat. This is probably simpler than what you want, though, and I don't believe it has a drag-and-drop-ish interface. That means you can't "draw" the forms, like you can with common GUI tools.
Another package, which is probably about as powerful as what you want, is newt (ships with RH and made by them, don't have an URL). The problem is that while this is a relatively high-level widget interface, it *still* isn't to the point of drawing your interface -- it requires coding. It's LGPLed.
May we never see th
That is, a continuous input stream that fills into the screens as they return data.
Go watch a few good travel agents sometime. They start their query, know what options are on successive screens, and are entering their choices before the screens even draw. Having seen those screens ten thousand times they have them memorized.
So, there might be 5 navigation steps to get to a desired screen but they don't see those 5 screens, they enter the navigation keypresses all while the second screen is drawing, which immediately is cleared as the next is drawn, etc, even if that stream is some weird string like 'R,N,1,3,F9'.
Now, one may argue that a good GUI would eliminate all those navigation steps by placing all the options on one, more dense, higher resolution, screen. Yet, in a complex system, not every element can be crammed on one screen, so there is still navigation that needs to be done, and the lack of type-ahead comes back to be a problem. Also, on an infinitely large screen the user takes an infinitely large amount of time to acquire his target, so every useful system of any complexity has to be multi-level.
There's also the issue of the keyboard being a much faster means of data entry than a mouse for certain types of data.
Plus, most terminals are of very low complexity and can be implemented to draw just as fast as their line speed allows. This compares favorably with, say, a web page full of DHTML, JS, CSS, and applets. Even a plain-old-html page often renders more slowly today because the browsers have to be capible of rendering so much more they're built on far more complex object models, which take time to process.
My God, it's Full of Source!
OUTSIDE_IP=$(dig +short my.ip @outsideip.net)