(which in the case of Wordpress can't be that hard)
(which in the case of Wordpress can't be that hard)
The core code is OK, security wise. Comparible to similar projects, I guess.
The two major issues are that typically PHP runs with permission to modify the WordPress directory, which is useful for automatic upgrades, but means any exploit in any plugin, theme, etc. instantly becomes, "can replace WordPress".
The other is that all themes and plugins are completely unsandboxed, and the quality is extremely variable.
The permissions thing can be fixed, by running a very stripped down wp-php user with only read access to the code, and only write access to wp-contents/uploads (and a logging dir). Then you do automatic upgrades with wp-cli and (real) cron from a user with write access.
The quality of plugins and themes is not easily fixed.
The API for writing plugins and themes doesn't help. It's archaic, spaghetti, and doesn't have any kind of coherenc. Global functions like wp_is_home(), the_loop(), get_sitename(), etc.
I've come across themes which bundle joomla inside them as they're really joomla themes with a shim.
Since when is implementing your own home-grown shitty replacement for parameterised queries "OK"?
> WordPress also works with PHP 5.2.4+ and MySQL 5.0+, but these versions have reached official End Of Life and as such may expose your site to security vulnerabilities
> PDO ships with PHP 5.1, and is available as a PECL extension for PHP 5.0
Unless you mean "hosts that disabled PDO"… I think they can safely be ignored.
Just checked, seems enabled by default on FreeBSD.
(Commented out in php.ini doesn't mean disabled — here it's enabled in separate files like /usr/local/etc/php/ext-30-pdo_mysql.ini)
MODX Revolution uses PDO exclusively, never had problems with it on any hosting service.
Commented out in php.ini does mean disabled - as in not enabled. PDO isn't enabled by default in FreeBSD you still need to install a separate package for PDO.
> Commented out in php.ini does mean disabled - as in not enabled.
Not necessarily, as I explained here: https://news.ycombinator.com/item?id=13761550
So while PDO may not be installed by a plain `apt-get install php5` or similar, I doubt the now deprecated `mysql` extension is installed by "default" in those scenarios either.
Edit: this approach also means that the PDO extension in php.ini will be commented, because its loaded by a package specific ini file e.g. /etc/php/7.0/fpm/conf.d/* which are generally symlinks to /etc/php/7.0/mods-available/
Here's a hint: Wordpress since v3.3 needs PHP 5.2+. MySQLi was added in PHP 5.0, PDO in 5.1
WordPress v3.3 was released in 2011.
This must be an oxymoron since complexity is the enemy of security.
I think what you meant to say was
> WordPress security is at best ad-hoc, and mostly non-existing
> show me if security is improved! > (which for Wordpress can't be that hard)
that is, to say that the concept of 'WordPress Security' isn't a single simple thing to 'improve'. Unfortunately. If you wrote a set of 10 jillion tests and proved that wordpress core has no possible security vulnerabilities (humour me), that still would not fix 'WordPress Security'.
From a theoretical point of view, you're absolutely right, and I agree with you.
From a practical point of view, WordPress is often 'good enough', and 'works well enough', and has 'few enough security issues', for many people.
Improving security is great, but it's a complex, perhaps impossible problem.
We're hosting a private-facing WordPress with several custom post types and some custom fields using ACF Pro. Then we pull directly out of the database with some not-too-complex and fast SQL queries and build a static site with Jekyll. The whole thing is about 200 lines of Ruby + a bunch of configuration for each post type. I'm working on making this a little more generic, because this thing really smokes.
In migrating to this, I'm doing a similar process with our previous mess of a WordPress site by pulling from the DB and writing JSON. WP All Import + ACF plugin takes care of the rest.
All uploads are now served out of S3 buckets and our editors write Kramdown now instead of HTML and inline CSS. We can easily A/B test or Multi-Armed Bandit now too.
It would have been a lot easier if the Wordpress REST API weren't a piece of junk. SQL is consistent at least.
Lastly, when pulling all the data, I determine if the post is HTML or Kramdown and use the Kramdown library to parse the HTML.
Everything migrated from the old site was automatically converted to Kramdown this same way and had to have some minor cleanup work done on each post. Not a small task on our main WordPress site with 7000+ posts/pages.
The interesting part was that wp_autop() happens at render time, so to get Kramdown to parse the HTML correctly, you actually have to RegExp replace '\n\n' with '<br /><br />' first to preserve paragraphs.
Most annoying though is that the API returns paginated posts but without returning data about how many posts there are or what page you're on. IIRC, there was a hard limit on returning max 100 posts per query. Might have been higher but significantly less than the number of posts we have. It was a total nonstarter unless we could guarantee that we could make and return enough queries to return all the posts before anything changed. With a 40+ content team that's impossible.
Sorry that I can't be more accurate than that right now.
Side note: I don't know if this has been said before, but ACF Pro is like the single most essential WordPress plugin for dev teams who know what they're doing. That plugin combined with some dev time replaces the need for like 99% of anything else that we would need to install. Please consider doing whatever it takes to make that plugin part of WP Core. Installing it on any site is a really sensible thing to do. That said, their field labels ("field_<somedigits>") really should be replaced with a GUID.
http://www.cvedetails.com/vulnerability-list/vendor_id-2337/...
WP security is like Whac-A-Mole, you patch one serious issue and two more pop up. It has been like this for years which shows that the developers don't have the correct mindset to write internet facing code.
I pointed you to a SQL injection / remote execution vulnerability in the WP core, in 2017 for gods sake.