The conclusion was to go ahead so I will commit the patch soon after
changing "glade" to "ui" as per Sven's comment.
In response to Akira's comment on performance:
We need to be aware of that we are introducing extra disk IO which can
give us problems, especialy during startup if we have to load a lot of
files. In practice I don't think this will be a problem for a long
time though, but doesn't hurt to have a mitigation strategy. We could:
* Put files as static strings in the code so that we don't have to
significant disk IO at all
* Merge several files together to minimize disk seeking times for many files
In response to the question on to what extent to use Glade:
We should not sacrifice good interaction design for simplicitly of
writing the UI code. UIs constructed through drag-and-drop of widgets
have a tendency to have bad usability, it is important that we keep
this in mind so we don't fall into the same trap.
In response to Torsten's comment on custom widgets:
Yes I agree, eventually we need to support our custom widgets with
Glade. This might require adaptions to our widgets and even to Glade and
In response to Michael:
"Less code" means less C code. In this case file-png.c now has 46 lines
of C code less. Also, you seemed to interpreted all my points as
arguments for using Glade while many of them were just comments. For
example, the bullet that we don't rely on any extra tool was just to
calm down people that don't want any new build dependency. The use of
Glade is optional but I except only extreme hackers to code the XML
files manually :).
My GIMP Blog:
"Best way to keep up with GIMP from git"
Gimp-developer mailing list