As of SS4.0.0 and the introduction of TrustedProxyMiddleware, the default now if no trusted proxies are defined is that nothing is a trusted proxy, whereas in SS3 a missing declaration was treated as everything being allowed.
* Remove overly specific PHP RNG instructions (that's just built into PHP7 through random_bytes now, which will throw if no suitable RNG is available)
* Remove PHP 5 RNG requirements, since we don't support that PHP release any mre
* Remove verbose explanation of PHP 5.6 support
* Remove conflicting instructions for PHP memory limits
* Remove version numbers from supporetd databases other than MySQL, it's up to the community modules to define that
* Remove Oracle support (code is nine years old!)
* Make "community supported" status clearer on databases, people can draw their own conclusions as open source users on Github
* Remove IIS version number, I think we should just stick to "needs web.config" and not give the impression that this is actively tested
* Remove mention of OSes for web servers, that's kind of irrelevant in today's hosting world (containers, PaaS, etc)
* Shorten install instructions in favour of a "quickstart" and point to lessons instead
* Remove mention of archive download option, we really shouldn't promote this - composer is the de-facto standard
* Add generic descriptions of the hosting environment considerations without going too much into specifics
* Remove Apache version number, we don't test on different versions, and really mostly rely on mod_rewrite working properly. Laravel does the same (doesn't claim specific Apache version support)
They don't account for protected files, or clear public webroot hosting.
I think this is more of a stackoverflow or forum.silverstripe.org article,
not something we should pretend to have official support for (or check for completeness and secure configuration)
We decided during implementation not to check permissions explicitly on cascading objects due to performance concerns.
For example, when publishing a page with embedded images, publish permissions on the image are implied - even if Image->canPublish() would return false for this author.
See https://github.com/silverstripe-security/security-issues/issues/57
