View Single Post
  #6   (View Single Post)  
Old 22nd August 2016
fvgit's Avatar
fvgit fvgit is offline
Spikes in tights
 
Join Date: May 2016
Location: perl -MMIME::Base64 -le 'print decode_base64("U2hlcndvb2QgRm9yZXN0")'
Posts: 314
Wink

Quote:
Originally Posted by shep View Post
There are two "flavors" of midori based on webkit: gtk2 and gtk3.
I already pulled a pkg_delete -X in the meantime. But to address your comment: pkg_add offered me to choose between midori-0.5.11 and midori-0.5.11-gtk3. I chose the former and ended up with gtk+3-3.18.7 AND gtk+2-2.24.29 anyway.

At first it coredumped everytime I tried to open a webpage. That is until I went through the preferences pane and discovered the proxy settings where I disabled the default gnome-something-proxy setting, don't remember what it was called exactly. After that I could fully load pages. But rendering was unbelievably slow and the browser still unstable. Mind you, my testbox consists of pretty old hardware (1998-2000), some might even consider it ancient.
Also, the whole dbus stuff and whatnot didn't sit very well with me.

BTW, the system has 320MB memory and the video card a whopping 4MB. And, yes, I do plan to test Firefox. Nothing like pushing the limits...

I might revisit midori at a later date. But on low-end hardware it is not a good choice.

Quote:
Originally Posted by TronDD View Post
Anything based on webkit is not going to be "light weight" in my opinion. It will be reasonably functional at rendering, though.

Light weight would be lynx, w3m, or links+ (my prefered).

Even dillo or netsurf will have fewer external dependencies.
Well, I consider text-mode browsing an entirely different kettle of fish. Somehow I never really got into it but I should probably revisit it. As for dillo you're spot on: here's a list of dillo's hand luggage on a system with zero packages installed:
Code:
quirks-2.197: ok
dillo-3.0.5p0:png-1.6.20: ok
dillo-3.0.5p0:libiconv-1.14p3: ok
dillo-3.0.5p0:gettext-0.19.7: ok
dillo-3.0.5p0:libidn-1.32: ok
dillo-3.0.5p0:libunistring-0.9.6: ok
dillo-3.0.5p0:libpsl-0.7.1p1: ok
dillo-3.0.5p0:pcre-8.38: ok
dillo-3.0.5p0:wget-1.16.3p0: ok
dillo-3.0.5p0:jpeg-9a: ok
dillo-3.0.5p0:fltk-1.3.3p0: ok
dillo-3.0.5p0:libelf-0.8.13p3: ok
dillo-3.0.5p0:libffi-3.2.1p0: ok
dillo-3.0.5p0:bzip2-1.0.6p7: ok
dillo-3.0.5p0:python-2.7.11: ok
dillo-3.0.5p0:glib2-2.46.2p0: ok
dillo-3.0.5p0:desktop-file-utils-0.22p0: ok
dillo-3.0.5p0: ok
Seems much more reasonable, doesn't it? And the best thing: it's pretty usable. It's not perfect, of course (what browser is?) but it works reasonably well within the feature set it supports. For dated hardware a pretty good choice.

Netsurf will be next on my list. I'm curious to see how it compares to dillo.

Quote:
Originally Posted by shep View Post
Browsers inherently have compromises - viewing embedded videos may be seen as both a convenience and a security risk. Running java-script (something netsurf and the text based browser are not capable of) is slower and provides ways to execute remote code on your system.
While I agree with everything you wrote, I'm still questioning the necessity of dependencies like cdparanoia for instance.
My point is, that somewhere along the road sth. went inherently wrong in matters of software design for web browsers. Especially if we consider the fact that basic functionality involves rendering a markup language document and a few images.
Let me put it his way: a web browser currently seems to be one the most un-OpenBSD-like pieces of software one might find on an OpenBSD desktop, text-mode variants notwithstanding.

Quote:
Originally Posted by jggimi View Post
1:20 for 83 pkg_adds might possibly be improved by using a closer mirror.
I wondered about that too, at first. I'm suspecting it might be partly hardware related (cable/nic). But for testing purposes I can live with it for the time being and look into it at a later date.
Reply With Quote