I really don't like that argument. Nobody is claiming that the Gnome dev's had to maintain GTK for everybody, or that it was their job. But it would be really nice, and it is disappointing that they don't.
For very long, the focus was ambiguous, and for some devs it was intended as a cross-platform toolkit (that's why there are Mac and Windows ports). Only "recently" (a few years) the focus has been on Gnome only.
> they could have easily forked GTK2 or 3 and maintained their own independent stable toolkit.
They could have, but almost nobody wants that. There is little utility in making yet another toolkit for a couple of new apps. What people want is to keep certain features in the mainline GTK, so mainline GTK apps retain them (non-CSD titlebars[1], icons in menus, stock actions, window-resize grippers, ...[2])
About code churn; I find the new versioning scheme of GTK particularly frustrating and backwards. Now they do have "stable" versions, but what they do is once they declare a version as stable, they break ABI, change the version number, and move all development to a new version. I think this is a useless definition of stable: People can already just throw the .so files in with their app if they want something frozen in time, there is no need for that kind of "stable" release.
What they used to do (in the late Gtk 2 days) was to keep backwards compatibility (within reasonable bounds) while adding new features. This is what I'd call stable - write code against 2.18, and it still runs against 2.28 a few years later, supporting new features and bugfixes. Now, that mode of development is harder, I understand, and they have no obligation to go that route. But it would be really beneficial if they want GTK to succeed. And I really want to like and use GTK (I prefer it and Gnome to some extent over Qt and KDE), this is why I get a bit worked up about the state of GTK...
[1] The titlebars are a bad example, AFAIK you can still get server side titlebars in the latest GTK.
[2] There are so many lost features, I actually made a list and at one point wanted to make a "Gnome Feature Graveyard" website in protest.
Except GTK is not the product, Gnome is. Your argument is absolutely true for end user products that are the focus of development. But the Gnome developers are not focused on GTK - they are focused on their desktop as their product, and the toolkit is just an abstraction to make development easier.
I think the schism was that this was a change of behavior, because like you said, GTK2 was pretty much toolkit first desktop second. The focus was on the toolkit that enabled the desktop, rather than the toolkit being along to complete the vision of the desktop. Back then I would definitely have considered GTK the premiere product of the Gnome developers.
A similar thing is happening, I would say, in reverse from KDE - long ago the applications and the desktop were the star, and the libraries were just there to enable the desktop. Then in KDE 5 the libraries were split up, dependencies were mapped, and documentation was improved - along with significant upstream contributions from KDE into qt5 to make that better - to make their libraries generic C++ libraries for general use. Today, I may not say the frameworks are the premier (Plasma Desktop and the applications suite are still huge) but they are at least equal in importance in terms of development effort - just consider how Kirigami and Plasma Mobile were announced - a library for development and a desktop component.
But it really should just demonstrate with these desktops its at the discretion of the developers themselves, rarely paid, volunteering their time, to spend it how they think it would be most valuable. If they don't move in the direction you want, you can be disappointing, but it is not the death of a product by misdirection - it is a realigning of objectives.
I would definitely not recommend anyone use GTK for anything but Gnome specific software within the Gnome project. Qt and the KDE Frameworks are just so good now its not even close in my mind, especially with versioning. But the Gnome developers still produce a fantastic desktop that I can easily showcase along side others and it shows its polish over alternatives like Plasma or Unity, and they make a valuable contribution for it.
You can mourn the loss of portable GTK, you can even suggest it be like that again, but in the face of a conscientious effort by its developers to deviate from that history, you have to respect that they are doing it intentionally and Internet complaints aren't going to change their minds, nor should they. I respect them for their contributions to free software too much to harp on them for not doing what I want.
For very long, the focus was ambiguous, and for some devs it was intended as a cross-platform toolkit (that's why there are Mac and Windows ports). Only "recently" (a few years) the focus has been on Gnome only.
> they could have easily forked GTK2 or 3 and maintained their own independent stable toolkit.
They could have, but almost nobody wants that. There is little utility in making yet another toolkit for a couple of new apps. What people want is to keep certain features in the mainline GTK, so mainline GTK apps retain them (non-CSD titlebars[1], icons in menus, stock actions, window-resize grippers, ...[2])
About code churn; I find the new versioning scheme of GTK particularly frustrating and backwards. Now they do have "stable" versions, but what they do is once they declare a version as stable, they break ABI, change the version number, and move all development to a new version. I think this is a useless definition of stable: People can already just throw the .so files in with their app if they want something frozen in time, there is no need for that kind of "stable" release.
What they used to do (in the late Gtk 2 days) was to keep backwards compatibility (within reasonable bounds) while adding new features. This is what I'd call stable - write code against 2.18, and it still runs against 2.28 a few years later, supporting new features and bugfixes. Now, that mode of development is harder, I understand, and they have no obligation to go that route. But it would be really beneficial if they want GTK to succeed. And I really want to like and use GTK (I prefer it and Gnome to some extent over Qt and KDE), this is why I get a bit worked up about the state of GTK...
[1] The titlebars are a bad example, AFAIK you can still get server side titlebars in the latest GTK. [2] There are so many lost features, I actually made a list and at one point wanted to make a "Gnome Feature Graveyard" website in protest.