diff --git a/docs/en/030_Supported_Protocols/005_Miniflux.md b/docs/en/030_Supported_Protocols/005_Miniflux.md
index 84e7d28..b29c082 100644
--- a/docs/en/030_Supported_Protocols/005_Miniflux.md
+++ b/docs/en/030_Supported_Protocols/005_Miniflux.md
@@ -10,7 +10,7 @@
     <dt>API endpoint</dt>
         <dd>/v1/</dd>
     <dt>Specifications</dt>
-        <dd><a href="https://miniflux.app/docs/api.html">API Reference</a></dd>
+        <dd><a href="https://miniflux.app/docs/api.html">API Reference</a>, <a href="https://miniflux.app/docs/rules.html#filtering-rules">Filtering Rules</a></dd>
 </dl>
 
 The Miniflux protocol is a fairly well-designed protocol supporting a wide variety of operations on newsfeeds, folders (termed "categories"), and articles; it also allows for user administration, and native OPML importing and exporting. Architecturally it is similar to the Nextcloud News protocol, but is generally more efficient and has more capabilities.
@@ -33,6 +33,13 @@ Miniflux version 2.0.26 is emulated, though not all features are implemented
 - Only the URL should be considered reliable in feed discovery results
 - The "All" category is treated specially (see below for details)
 - Category names consisting only of whitespace are rejected along with the empty string
+- Filtering rules may not function identically (see below for details)
+
+# Behaviour of filtering (block and keep) rules
+
+The Miniflux documentation gives only a brief example of a pattern for its filtering rules; the allowed syntax is described in full [in Google's documentation for RE2](https://github.com/google/re2/wiki/Syntax). Being a PHP application, The Arsse instead accepts [PCRE syntax](http://www.pcre.org/original/doc/html/pcresyntax.html) (or since PHP 7.3 [PCRE2 syntax](https://www.pcre.org/current/doc/html/pcre2syntax.html)), specifically in UTF-8 mode. Delimiters should not be included, and slashes should not be escaped; anchors may be used if desired. For example `^(?i)RE/MAX$` is a valid pattern.
+
+For convenience the patterns are tested after collapsing whitespace. Unlike Miniflux, The Arsse tests the patterns against an article's author-supplied categories if they do not match its title.
 
 # Special handling of the "All" category
 
diff --git a/sql/MySQL/6.sql b/sql/MySQL/6.sql
index 36b2d6e..9370e27 100644
--- a/sql/MySQL/6.sql
+++ b/sql/MySQL/6.sql
@@ -6,6 +6,9 @@
 
 alter table arsse_tokens add column data longtext default null;
 
+alter table arsse_subscriptions add column keep_rule longtext default null;
+alter table arsse_subscriptions add column block_rule longtext default null;
+
 alter table arsse_users add column num bigint unsigned unique;
 alter table arsse_users add column admin boolean not null default 0;
 create temporary table arsse_users_existing(
diff --git a/sql/PostgreSQL/6.sql b/sql/PostgreSQL/6.sql
index a32eb0c..f936b87 100644
--- a/sql/PostgreSQL/6.sql
+++ b/sql/PostgreSQL/6.sql
@@ -6,6 +6,9 @@
 
 alter table arsse_tokens add column data text default null;
 
+alter table arsse_subscriptions add column keep_rule text default null;
+alter table arsse_subscriptions add column block_rule text default null;
+
 alter table arsse_users add column num bigint unique;
 alter table arsse_users add column admin smallint not null default 0;
 create temp table arsse_users_existing(
diff --git a/sql/SQLite3/6.sql b/sql/SQLite3/6.sql
index 81e9e82..752c056 100644
--- a/sql/SQLite3/6.sql
+++ b/sql/SQLite3/6.sql
@@ -6,8 +6,11 @@
 -- This is a speculative addition to support OAuth login in the future
 alter table arsse_tokens add column data text default null;
 
--- Add num and admin columns to the users table
--- In particular this adds a numeric identifier for each user, which Miniflux requires
+-- Add columns to subscriptions to store "keep" and "block" filtering rules from Miniflux
+alter table arsse_subscriptions add column keep_rule text default null;
+alter table arsse_subscriptions add column block_rule text default null;
+
+-- Add numeric identifier and admin columns to the users table
 create table arsse_users_new(
 -- users
     id text primary key not null collate nocase,    -- user id