Commit 1c0129c7 authored by Gustavo Noronha Silva's avatar Gustavo Noronha Silva Committed by Emanuele Aina

webkit-clutter -> web-engine: long overdue update

The document was 2 major changes out of date. WebKit Clutter has not
been used in Apertis for a few years. More recently, a shift in focus
has changed the ammount of effort the Apertis project is willing to
dedicate to maintaining custom changes, which means a move to upstream.
Signed-off-by: 's avatarGustavo Noronha Silva <gustavo.noronha@collabora.com>
parent 981c0aee
---
title: Web engine
short-description: Notes on the web engine provided by Apertis
authors:
- name: Gustavo Noronha Silva
...
# Web engine
## Introduction
Apertis provides the GTK port of WebKit as its web engine. To
ensure low maintenance effort, no changes are made to the downstream
branch, any improvements should go to the upstream project.
### Security maintenance
Like all other major browser engines, the GTK port does not provide long
term support, so security maintenance comes down to staying up to date.
The general approach Apertis takes is to follow whatever Debian provides.
The project may also importing a new upstream release that has not been
made available in Debian yet if an important fix is available.
### Customization
Apertis has made a decision to not perform customizations to the engine
with the goal of keeping maintenance efforts to a minimum. Whenever a
feature is desired, it should be proposed and contributed directly
upstream.
#### White-listing and black-listing
There has been interest in maintaining a black-list of web applications
(or pages) that misbehaved. That would be for the case in which the
browser gets killed because it stopped responding and the scripts
watchdog was not able to restore it to working, so that those web apps
or pages are not loaded automatically upon startup causing the browser
to go unresponsive again.
[Web] \(codename [Epiphany]), the GNOME web browser maintains a
session file that stores information about all loaded pages, such as
title, URL, and whether they are currently loading or not. If Web is
quit unexpectedly, it will refuse to load any pages that were marked as
still loading automatically. This same approach could be used by the
Apertis web browser to not load those pages automatically or to create a
blacklist database.
The white-list, on the other hand, would be used to enable applications
to use lots of resources for a long time without getting killed by this
infrastructure in what could be considered a false positive. A
white-list can easily be implemented, keeping a list of applications
that are allowed to go over the limits should be enough.
#### Rendering of non-web documents
Several kinds of documents that are not strictly web documents are
available on web sites for viewing and download. Some of these types of
documents, such as PDFs, have become so common that some browsers embed
a viewer.
WebKit itself does not have support for rendering those documents and
the WebView widget provided by WebKitGTK does not support any
kind of custom rendering. Applications and browsers that use the engine
can also embed different widget alongside the WebView if they would like
to allow viewing PDFs and other kinds of documents on the same user
interface.
## Scheduled and potential future work
### Web runtime
There is interest in providing developers with a way to write
applications using web technologies for Apertis. While this is out of
the scope of this design, a small description of existing technologies
and how they can be applied follows. Collabora can help in the future
with a more detailed analysis of what works needs doing, specification
and development of a solution.
A runner for web applications would ideally create a process for each
separate application that will be executed, and use application-specific
locations for storing data such as caches and the databases for the
features described above – meaning it would not require any kind of
special privilege, it would be a regular application. This also means
permissions and resource limits can be set individually also for web
applications.
The fact that more than one process would be executed does not mean a
lot of memory overhead, since shared libraries are only loaded in memory
once for all processes that use them. This would also have several
advantages such as making managing applications permissions easier, and
avoiding one application interfering with others.
Collabora believes that the best way to provide platform APIs to the
web applications would be through custom bindings for relevant interfaces.
That ensures the APIs will feel native to the JavaScript environment,
and will significantly reduce the attack surface presented to pages,
compared to a solution that makes all native libraries available.
[Web]: https://live.gnome.org/Design/Web
[Epiphany]: https://live.gnome.org/Epiphany
This diff is collapsed.
......@@ -31,6 +31,5 @@ index.md
system-updates-and-rollback.md
text-to-speech.md
ui-customisation.md
webkit-clutter.md
web-engine.md
web-portal-caching.md
web-runtime.md
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment