2 deleted 3 characters in body
source | link

I'd like to see such a library which is elegant and simple to use with Haskell. The rest is technical details that should serve this purpose, not redefine it. SoThus my $0.02.

Don't base it on an existing toolkit, like Qt or GTK or FLTK or... — this will severely limit you and will probably give you far more pain than profit. PyQt is, erm, funny and contrived enough, and both Python and C++ are extremely flexible imperative OO languages. In the case of Haskell, things would be much rougher, I suppose.

Depend only on the most basic graphic primitives, then build upon that. OpenGL is nice, but even something simpler (2D-only, e.g. SDL) would do well, too. This will give you maximum flexibility and maximum portability. See Smalltalk/Morphic, Java/Swing, TCL/Tk.

Make it conceptually small. GUIs are hard as they are, no need to add another Everest to climb. Haskell, I hope, can help make the thing compact and modular.

For bonus points, make it skinnable. As a minimum, know how to apply system colors (and only system colors) to paint your entire repertoire of controls, so that an app built with this toolkit is not a eyesore. As a maximum, know how to make Win32/Gtk/Qt/Cocoa draw your controls so that they look fully native. Basic skinnability is simple and logical; attaining full native look is quite hard.

Also, please run rootless and leave window management to the underlying graphic system — X, Windows, whatever. Not doing so will challenge users' sanity and hamper adoption drastically.

As usual, 'make simple things simple and complex things possible' + 'avoid the Turing tarpit where everything is possible but nothing of interest is simple' + 'make at as simple as possible but not simpler'.

The name is the least important thing. Of all popular GUI toolkits, only Qt has a somehow clever name. A number of popular projects changed names, even in-flight (Firefox, née Firebird). Have something to name, and you'll name it.

Good luck!

I'd like to see such a library which is elegant and simple to use with Haskell. The rest is technical details that should serve this purpose, not redefine it. So my $0.02.

Don't base it on an existing toolkit, like Qt or GTK or FLTK or... — this will severely limit you and will probably give you far more pain than profit. PyQt is, erm, funny and contrived enough, and both Python and C++ are extremely flexible imperative OO languages. In the case of Haskell, things would be much rougher, I suppose.

Depend only on the most basic graphic primitives, then build upon that. OpenGL is nice, but even something simpler (2D-only, e.g. SDL) would do well, too. This will give you maximum flexibility and maximum portability. See Smalltalk/Morphic, Java/Swing, TCL/Tk.

Make it conceptually small. GUIs are hard as they are, no need to add another Everest to climb. Haskell, I hope, can help make the thing compact and modular.

For bonus points, make it skinnable. As a minimum, know how to apply system colors (and only system colors) to paint your entire repertoire of controls, so that an app built with this toolkit is not a eyesore. As a maximum, know how to make Win32/Gtk/Qt/Cocoa draw your controls so that they look fully native. Basic skinnability is simple and logical; attaining full native look is quite hard.

Also, please run rootless and leave window management to the underlying graphic system — X, Windows, whatever. Not doing so will challenge users' sanity and hamper adoption drastically.

As usual, 'make simple things simple and complex things possible' + 'avoid the Turing tarpit where everything is possible but nothing of interest is simple' + 'make at as simple as possible but not simpler'.

The name is the least important thing. Of all popular GUI toolkits, only Qt has a somehow clever name. A number of popular projects changed names, even in-flight (Firefox, née Firebird). Have something to name, and you'll name it.

Good luck!

I'd like to see a library which is elegant and simple to use with Haskell. The rest is technical details that should serve this purpose, not redefine it. Thus my $0.02.

Don't base it on an existing toolkit, like Qt or GTK or FLTK or... — this will severely limit you and will probably give you far more pain than profit. PyQt is, erm, funny and contrived enough, and both Python and C++ are extremely flexible imperative OO languages. In the case of Haskell, things would be much rougher, I suppose.

Depend only on the most basic graphic primitives, then build upon that. OpenGL is nice, but even something simpler (2D-only, e.g. SDL) would do well, too. This will give you maximum flexibility and maximum portability. See Smalltalk/Morphic, Java/Swing, TCL/Tk.

Make it conceptually small. GUIs are hard as they are, no need to add another Everest to climb. Haskell, I hope, can help make the thing compact and modular.

For bonus points, make it skinnable. As a minimum, know how to apply system colors (and only system colors) to paint your entire repertoire of controls, so that an app built with this toolkit is not a eyesore. As a maximum, know how to make Win32/Gtk/Qt/Cocoa draw your controls so that they look fully native. Basic skinnability is simple and logical; attaining full native look is quite hard.

Also, please run rootless and leave window management to the underlying graphic system — X, Windows, whatever. Not doing so will challenge users' sanity and hamper adoption drastically.

As usual, 'make simple things simple and complex things possible' + 'avoid the Turing tarpit where everything is possible but nothing of interest is simple' + 'make at as simple as possible but not simpler'.

The name is the least important thing. Of all popular GUI toolkits, only Qt has a somehow clever name. A number of popular projects changed names, even in-flight (Firefox, née Firebird). Have something to name, and you'll name it.

Good luck!

1
source | link

I'd like to see such a library which is elegant and simple to use with Haskell. The rest is technical details that should serve this purpose, not redefine it. So my $0.02.

Don't base it on an existing toolkit, like Qt or GTK or FLTK or... — this will severely limit you and will probably give you far more pain than profit. PyQt is, erm, funny and contrived enough, and both Python and C++ are extremely flexible imperative OO languages. In the case of Haskell, things would be much rougher, I suppose.

Depend only on the most basic graphic primitives, then build upon that. OpenGL is nice, but even something simpler (2D-only, e.g. SDL) would do well, too. This will give you maximum flexibility and maximum portability. See Smalltalk/Morphic, Java/Swing, TCL/Tk.

Make it conceptually small. GUIs are hard as they are, no need to add another Everest to climb. Haskell, I hope, can help make the thing compact and modular.

For bonus points, make it skinnable. As a minimum, know how to apply system colors (and only system colors) to paint your entire repertoire of controls, so that an app built with this toolkit is not a eyesore. As a maximum, know how to make Win32/Gtk/Qt/Cocoa draw your controls so that they look fully native. Basic skinnability is simple and logical; attaining full native look is quite hard.

Also, please run rootless and leave window management to the underlying graphic system — X, Windows, whatever. Not doing so will challenge users' sanity and hamper adoption drastically.

As usual, 'make simple things simple and complex things possible' + 'avoid the Turing tarpit where everything is possible but nothing of interest is simple' + 'make at as simple as possible but not simpler'.

The name is the least important thing. Of all popular GUI toolkits, only Qt has a somehow clever name. A number of popular projects changed names, even in-flight (Firefox, née Firebird). Have something to name, and you'll name it.

Good luck!