On Fri, Feb 24, 2012 at 4:17 PM, Jeffrey Van Voorst <[email protected]> wrote: > >From my point of view, it seems that I am still missing a list of > dependencies for pythonwebkit. As from previous comments, I infer I > don't need the gobject-introspection package to get pyjamas desktop > working, but from the different notes I read I had inferred that > pythonwebkit requires the python bindings for GTK+. I may also be > confusing pywebkit and pythonwebkit.
yes, you are. pywebkitgtk originally treated webkit as a sort of "hands-off" black box. you could set the URL, get notification about back and forward events, and progress notification, but that was about it. *no* DOM bindings *at all*. in 2008 or so i added the gobject bindings, and the natural place to ground them was in pywebkitgtk. pywebkitgtk therefore went up from a 180k library to a whopping 2mb one. i'd designed the code to use one signal with sub-signals, and to activate them "on-demand", but i had *not* realised the impact of using gobject_timer at the time. so the 2nd attempt i created "independent" - direct - python bindings - that could be added to absolutely any webkit "port". i proved the point by adding the exact same python bindings to *both* the webkit-gtk port *and* the (new) webkit-directfb port. i also tried adding them to the webkit-qt4 port but couldn't get to grips with the build infrastructure (qmake). it's a bit... obtuse :) now it looks like people are finally understanding the problems associated with the original design of the gobject bindings, and they might just be reversing some of the immense damage that mark rowe did. > At this point, I will probably just try to build webkit from git clone > git://git.savannah.gnu.org/pythonwebkit.git after checking out the > python_codegen branch. I will see how far I can get by installing > dependencies based on compile/other errors. you can shortcut that by using the advice on the page [apt-get build-dep libwebkitgtk and other tricks] > On a completely different note, there are some comments floating > around "teh interwebs" about glib's event loop being painfully slow. > For that reason, any project using glib's event loop should do as much > as humanly possible to not add unnecessary events (which it appears > pyjamas desktop does, but some other solutions do not). hmmm... do you happen to know what those solutions are? the only reason for using glib's event loop is because i did not know of the existence of any alternatives: it's not a critical requirement. in the python_codegen DirectFB port, the directfb GUI equivalent is used. i don't know its performance characteristics. > In a sense, what I am trying to do is figure out where to best spend > my limited time, as I don't have the desire or time to get into a > pissing match with core developers of a desktop or browser engine. neither do i. however neither do i wish people to forget that there exist people who like to be bullies - who enjoy "teaching people a lesson" regardless of the consequences. the consequences in this case are that the introduction of useful and useable bindings to DOM in webkit which are suitable for use by the free software community has been delayed by almost *four* years. l.
