diff --git a/pillar/api/activities.py b/pillar/api/activities.py
index b4bffa34..107377e3 100644
--- a/pillar/api/activities.py
+++ b/pillar/api/activities.py
@@ -119,6 +119,8 @@ def activity_subscribe(user_id, context_object_type, context_object_id):
 
     # If no subscription exists, we create one
     if not subscription:
+        # Workaround for issue: https://github.com/pyeve/eve/issues/1174
+        lookup['notifications'] = {}
         current_app.post_internal('activities-subscriptions', lookup)
 
 
diff --git a/pillar/api/custom_field_validation.py b/pillar/api/custom_field_validation.py
index 82a51730..4796676f 100644
--- a/pillar/api/custom_field_validation.py
+++ b/pillar/api/custom_field_validation.py
@@ -1,4 +1,3 @@
-import copy
 from datetime import datetime
 import logging
 
@@ -6,36 +5,12 @@ from bson import ObjectId, tz_util
 from eve.io.mongo import Validator
 from flask import current_app
 
-import pillar.markdown
+from pillar import markdown
 
 log = logging.getLogger(__name__)
 
 
 class ValidateCustomFields(Validator):
-    def __init__(self, *args, **kwargs):
-        super().__init__(*args, **kwargs)
-
-        # Will be reference to the actual document being validated, so that we can
-        # modify it during validation.
-        self.__real_document = None
-
-    def validate(self, document, *args, **kwargs):
-        # Keep a reference to the actual document, because Cerberus validates copies.
-        self.__real_document = document
-        result = super().validate(document, *args, **kwargs)
-
-        # Store the in-place modified document as self.document, so that Eve's post_internal
-        # can actually pick it up as the validated document. We need to make a copy so that
-        # further modifications (like setting '_etag' etc.) aren't done in-place.
-        self.document = copy.deepcopy(document)
-
-        return result
-
-    def _get_child_validator(self, *args, **kwargs):
-        child = super()._get_child_validator(*args, **kwargs)
-        # Pass along our reference to the actual document.
-        child.__real_document = self.__real_document
-        return child
 
     # TODO: split this into a convert_property(property, schema) and call that from this function.
     def convert_properties(self, properties, node_schema):
@@ -137,8 +112,7 @@ class ValidateCustomFields(Validator):
         if val:
             # This ensures the modifications made by v's coercion rules are
             # visible to this validator's output.
-            # TODO(fsiddi): this no longer works due to Cerberus internal changes.
-            # self.current[field] = v.current
+            self.document[field] = v.document
             return True
 
         log.warning('Error validating properties for node %s: %s', self.document, v.errors)
@@ -183,36 +157,18 @@ class ValidateCustomFields(Validator):
         if ip.prefixlen() == 0:
             self._error(field_name, 'Zero-length prefix is not allowed')
 
-    def _validator_markdown(self, field, value):
-        """Convert MarkDown.
+    def _normalize_coerce_markdown(self, markdown_field: str) -> str:
         """
-        my_log = log.getChild('_validator_markdown')
+        Cache markdown as html.
 
-        # Find this field inside the original document
-        my_subdoc = self._subdoc_in_real_document()
-        if my_subdoc is None:
-            # If self.update==True we are validating an update document, which
-            # may not contain all fields, so then a missing field is fine.
-            if not self.update:
-                self._error(field, f'validator_markdown: unable to find sub-document '
-                                   f'for path {self.document_path}')
-            return
-
-        my_log.debug('validating field %r with value %r', field, value)
-        save_to = pillar.markdown.cache_field_name(field)
-        html = pillar.markdown.markdown(value)
-        my_log.debug('saving result to %r in doc with id %s', save_to, id(my_subdoc))
-        my_subdoc[save_to] = html
-
-    def _subdoc_in_real_document(self):
-        """Return a reference to the current sub-document inside the real document.
-
-        This allows modification of the document being validated.
+        :param markdown_field: name of the field containing mark down
+        :return: html string
         """
-        my_subdoc = getattr(self, 'persisted_document') or self.__real_document
-        for item in self.document_path:
-            my_subdoc = my_subdoc[item]
-        return my_subdoc
+        my_log = log.getChild('_normalize_coerce_markdown')
+        mdown = self.document.get(markdown_field, '')
+        html = markdown.markdown(mdown)
+        my_log.debug('Generated html for markdown field %s in doc with id %s', markdown_field, id(self.document))
+        return html
 
 
 if __name__ == '__main__':
diff --git a/pillar/api/eve_settings.py b/pillar/api/eve_settings.py
index c980066d..82d3b441 100644
--- a/pillar/api/eve_settings.py
+++ b/pillar/api/eve_settings.py
@@ -1,5 +1,7 @@
 import os
 
+from pillar.api.node_types.utils import markdown_fields
+
 STORAGE_BACKENDS = ["local", "pillar", "cdnsun", "gcs", "unittest"]
 URL_PREFIX = 'api'
 
@@ -184,12 +186,7 @@ organizations_schema = {
         'maxlength': 128,
         'required': True
     },
-    'description': {
-        'type': 'string',
-        'maxlength': 256,
-        'validator': 'markdown',
-    },
-    '_description_html': {'type': 'string'},
+    **markdown_fields('description', maxlength=256),
     'website': {
         'type': 'string',
         'maxlength': 256,
@@ -322,11 +319,7 @@ nodes_schema = {
         'maxlength': 128,
         'required': True,
     },
-    'description': {
-        'type': 'string',
-        'validator': 'markdown',
-    },
-    '_description_html': {'type': 'string'},
+    **markdown_fields('description'),
     'picture': _file_embedded_schema,
     'order': {
         'type': 'integer',
@@ -576,11 +569,7 @@ projects_schema = {
         'maxlength': 128,
         'required': True,
     },
-    'description': {
-        'type': 'string',
-        'validator': 'markdown',
-    },
-    '_description_html': {'type': 'string'},
+    **markdown_fields('description'),
     # Short summary for the project
     'summary': {
         'type': 'string',
diff --git a/pillar/api/node_types/__init__.py b/pillar/api/node_types/__init__.py
index 84b0e0ac..4a25eb40 100644
--- a/pillar/api/node_types/__init__.py
+++ b/pillar/api/node_types/__init__.py
@@ -23,14 +23,6 @@ attachments_embedded_schema = {
                 'type': 'objectid',
                 'required': True,
             },
-            'link': {
-                'type': 'string',
-                'allowed': ['self', 'none', 'custom'],
-                'default': 'self',
-            },
-            'link_custom': {
-                'type': 'string',
-            },
             'collection': {
                 'type': 'string',
                 'allowed': ['files'],
diff --git a/pillar/api/node_types/comment.py b/pillar/api/node_types/comment.py
index e12b3bab..d98a1cd8 100644
--- a/pillar/api/node_types/comment.py
+++ b/pillar/api/node_types/comment.py
@@ -1,17 +1,15 @@
 from pillar.api.node_types import attachments_embedded_schema
+from pillar.api.node_types.utils import markdown_fields
 
 node_type_comment = {
     'name': 'comment',
     'description': 'Comments for asset nodes, pages, etc.',
     'dyn_schema': {
         # The actual comment content
-        'content': {
-            'type': 'string',
-            'minlength': 5,
-            'required': True,
-            'validator': 'markdown',
-        },
-        '_content_html': {'type': 'string'},
+        **markdown_fields(
+            'content',
+            minlength=5,
+            required=True),
         'status': {
             'type': 'string',
             'allowed': [
diff --git a/pillar/api/node_types/post.py b/pillar/api/node_types/post.py
index c5afd24e..71529236 100644
--- a/pillar/api/node_types/post.py
+++ b/pillar/api/node_types/post.py
@@ -1,17 +1,14 @@
 from pillar.api.node_types import attachments_embedded_schema
+from pillar.api.node_types.utils import markdown_fields
 
 node_type_post = {
     'name': 'post',
     'description': 'A blog post, for any project',
     'dyn_schema': {
-        'content': {
-            'type': 'string',
-            'minlength': 5,
-            'maxlength': 90000,
-            'required': True,
-            'validator': 'markdown',
-        },
-        '_content_html': {'type': 'string'},
+        **markdown_fields('content',
+                          minlength=5,
+                          maxlength=90000,
+                          required=True),
         'status': {
             'type': 'string',
             'allowed': [
diff --git a/pillar/api/node_types/utils.py b/pillar/api/node_types/utils.py
new file mode 100644
index 00000000..73c7d4a2
--- /dev/null
+++ b/pillar/api/node_types/utils.py
@@ -0,0 +1,34 @@
+from pillar import markdown
+
+
+def markdown_fields(field: str, **kwargs) -> dict:
+    """
+    Creates a field for the markdown, and a field for the cached html.
+
+    Example usage:
+    schema = {'myDoc': {
+        'type': 'list',
+        'schema': {
+            'type': 'dict',
+            'schema': {
+                **markdown_fields('content', required=True),
+            }
+        },
+    }}
+
+    :param field:
+    :return:
+    """
+    cache_field = markdown.cache_field_name(field)
+    return {
+        field: {
+            'type': 'string',
+            **kwargs
+        },
+        cache_field: {
+            'type': 'string',
+            'readonly': True,
+            'default': field,  # Name of the field containing the markdown. Will be input to the coerce function.
+            'coerce': 'markdown',
+        }
+    }
\ No newline at end of file
diff --git a/pillar/api/nodes/__init__.py b/pillar/api/nodes/__init__.py
index c3f76389..60b36881 100644
--- a/pillar/api/nodes/__init__.py
+++ b/pillar/api/nodes/__init__.py
@@ -247,14 +247,12 @@ def setup_app(app, url_prefix):
     app.on_fetched_resource_nodes += eve_hooks.before_returning_nodes
 
     app.on_replace_nodes += eve_hooks.before_replacing_node
-    app.on_replace_nodes += eve_hooks.parse_markdown
     app.on_replace_nodes += eve_hooks.texture_sort_files
     app.on_replace_nodes += eve_hooks.deduct_content_type_and_duration
     app.on_replace_nodes += eve_hooks.node_set_default_picture
     app.on_replaced_nodes += eve_hooks.after_replacing_node
 
     app.on_insert_nodes += eve_hooks.before_inserting_nodes
-    app.on_insert_nodes += eve_hooks.parse_markdowns
     app.on_insert_nodes += eve_hooks.nodes_deduct_content_type_and_duration
     app.on_insert_nodes += eve_hooks.nodes_set_default_picture
     app.on_insert_nodes += eve_hooks.textures_sort_files
diff --git a/pillar/api/nodes/eve_hooks.py b/pillar/api/nodes/eve_hooks.py
index ba131f87..891f4a11 100644
--- a/pillar/api/nodes/eve_hooks.py
+++ b/pillar/api/nodes/eve_hooks.py
@@ -122,47 +122,50 @@ def before_inserting_nodes(items):
         # Default the 'user' property to the current user.
         item.setdefault('user', current_user.user_id)
 
+def get_comment_verb_and_context_object_id(comment):
+    nodes_collection = current_app.data.driver.db['nodes']
+    verb = 'commented'
+    parent = nodes_collection.find_one({'_id': comment['parent']})
+    context_object_id = comment['parent']
+    while parent['node_type'] == 'comment':
+        # If the parent is a comment, we provide its own parent as
+        # context. We do this in order to point the user to an asset
+        # or group when viewing the notification.
+        verb = 'replied'
+        context_object_id = parent['parent']
+        parent = nodes_collection.find_one({'_id': parent['parent']})
+    return verb, context_object_id
+
 
 def after_inserting_nodes(items):
     for item in items:
-        # Skip subscriptions for first level items (since the context is not a
-        # node, but a project).
+        context_object_id = None
         # TODO: support should be added for mixed context
-        if 'parent' not in item:
-            return
-        context_object_id = item['parent']
-        if item['node_type'] == 'comment':
-            nodes_collection = current_app.data.driver.db['nodes']
-            parent = nodes_collection.find_one({'_id': item['parent']})
-            # Always subscribe to the parent node
-            activity_subscribe(item['user'], 'node', item['parent'])
-            if parent['node_type'] == 'comment':
-                # If the parent is a comment, we provide its own parent as
-                # context. We do this in order to point the user to an asset
-                # or group when viewing the notification.
-                verb = 'replied'
-                context_object_id = parent['parent']
-                # Subscribe to the parent of the parent comment (post or group)
-                activity_subscribe(item['user'], 'node', parent['parent'])
-            else:
-                activity_subscribe(item['user'], 'node', item['_id'])
-                verb = 'commented'
-        elif item['node_type'] in PILLAR_NAMED_NODE_TYPES:
-            verb = 'posted'
+        if item['node_type'] in PILLAR_NAMED_NODE_TYPES:
             activity_subscribe(item['user'], 'node', item['_id'])
-        else:
-            # Don't automatically create activities for non-Pillar node types,
-            # as we don't know what would be a suitable verb (among other things).
-            continue
+            verb = 'posted'
+            context_object_id = item.get('parent')
+            if item['node_type'] == 'comment':
+                # Always subscribe to the parent node
+                activity_subscribe(item['user'], 'node', item['parent'])
+                verb, context_object_id = get_comment_verb_and_context_object_id(item)
+                # Subscribe to the parent of the parent comment (post or group)
+                activity_subscribe(item['user'], 'node', context_object_id)
 
-        activity_object_add(
-            item['user'],
-            verb,
-            'node',
-            item['_id'],
-            'node',
-            context_object_id
-        )
+
+        if context_object_id and item['node_type'] in PILLAR_NAMED_NODE_TYPES:
+            # * Skip activity for first level items (since the context is not a
+            # node, but a project).
+            # * Don't automatically create activities for non-Pillar node types,
+            # as we don't know what would be a suitable verb (among other things).
+            activity_object_add(
+                item['user'],
+                verb,
+                'node',
+                item['_id'],
+                'node',
+                context_object_id
+            )
 
 
 def deduct_content_type_and_duration(node_doc, original=None):
@@ -322,46 +325,6 @@ def textures_sort_files(nodes):
         texture_sort_files(node)
 
 
-def parse_markdown(node, original=None):
-    import copy
-
-    projects_collection = current_app.data.driver.db['projects']
-    project = projects_collection.find_one({'_id': node['project']}, {'node_types': 1})
-    # Query node type directly using the key
-    node_type = next(nt for nt in project['node_types']
-                     if nt['name'] == node['node_type'])
-
-    # Create a copy to not overwrite the actual schema.
-    schema = copy.deepcopy(current_app.config['DOMAIN']['nodes']['schema'])
-    schema['properties'] = node_type['dyn_schema']
-
-    def find_markdown_fields(schema, node):
-        """Find and process all makrdown validated fields."""
-        for k, v in schema.items():
-            if not isinstance(v, dict):
-                continue
-
-            if v.get('validator') == 'markdown':
-                # If there is a match with the validator: markdown pair, assign the sibling
-                # property (following the naming convention _<property>_html)
-                # the processed value.
-                if k in node:
-                    html = pillar.markdown.markdown(node[k])
-                    field_name = pillar.markdown.cache_field_name(k)
-                    node[field_name] = html
-            if isinstance(node, dict) and k in node:
-                find_markdown_fields(v, node[k])
-
-    find_markdown_fields(schema, node)
-
-    return 'ok'
-
-
-def parse_markdowns(items):
-    for item in items:
-        parse_markdown(item)
-
-
 def short_link_info(short_code):
     """Returns the short link info in a dict."""
 
diff --git a/pillar/api/utils/__init__.py b/pillar/api/utils/__init__.py
index df279287..c244ba63 100644
--- a/pillar/api/utils/__init__.py
+++ b/pillar/api/utils/__init__.py
@@ -44,10 +44,16 @@ def remove_private_keys(document):
     """Removes any key that starts with an underscore, returns result as new
     dictionary.
     """
+    def do_remove(doc):
+        for key in list(doc.keys()):
+            if key.startswith('_'):
+                del doc[key]
+            elif isinstance(doc[key], dict):
+                doc[key] = do_remove(doc[key])
+        return doc
+
     doc_copy = copy.deepcopy(document)
-    for key in list(doc_copy.keys()):
-        if key.startswith('_'):
-            del doc_copy[key]
+    do_remove(doc_copy)
 
     try:
         del doc_copy['allowed_methods']
diff --git a/pillar/cli/maintenance.py b/pillar/cli/maintenance.py
index 6ad3c6f5..8dc873ae 100644
--- a/pillar/cli/maintenance.py
+++ b/pillar/cli/maintenance.py
@@ -739,113 +739,6 @@ def iter_markdown(proj_node_types: dict, some_node: dict, callback: typing.Calla
             doc[key] = new_value
 
 
-@manager_maintenance.option('-p', '--project', dest='proj_url', nargs='?',
-                            help='Project URL')
-@manager_maintenance.option('-a', '--all', dest='all_projects', action='store_true', default=False,
-                            help='Replace on all projects.')
-@manager_maintenance.option('-g', '--go', dest='go', action='store_true', default=False,
-                            help='Actually perform the changes (otherwise just show as dry-run).')
-def upgrade_attachment_usage(proj_url=None, all_projects=False, go=False):
-    """Replaces '@[slug]' with '{attachment slug}'.
-
-    Also moves links from the attachment dict to the attachment shortcode.
-    """
-    if bool(proj_url) == all_projects:
-        log.error('Use either --project or --all.')
-        return 1
-
-    import html
-    from pillar.api.projects.utils import node_type_dict
-    from pillar.api.utils import remove_private_keys
-    from pillar.api.utils.authentication import force_cli_user
-
-    force_cli_user()
-
-    nodes_coll = current_app.db('nodes')
-    total_nodes = 0
-    failed_node_ids = set()
-
-    # Use a mixture of the old slug RE that still allowes spaces in the slug
-    # name and the new RE that allows dashes.
-    old_slug_re = re.compile(r'@\[([a-zA-Z0-9_\- ]+)\]')
-    for proj in _db_projects(proj_url, all_projects, go=go):
-        proj_id = proj['_id']
-        proj_url = proj.get('url', '-no-url-')
-        nodes = nodes_coll.find({
-            '_deleted': {'$ne': True},
-            'project': proj_id,
-            'properties.attachments': {'$exists': True},
-        })
-        node_count = nodes.count()
-        if node_count == 0:
-            log.debug('Skipping project %s (%s)', proj_url, proj_id)
-            continue
-
-        proj_node_types = node_type_dict(proj)
-
-        for node in nodes:
-            attachments = node['properties']['attachments']
-            replaced = False
-
-            # Inner functions because of access to the node's attachments.
-            def replace(match):
-                nonlocal replaced
-                slug = match.group(1)
-                log.debug('    - OLD STYLE attachment slug %r', slug)
-                try:
-                    att = attachments[slug]
-                except KeyError:
-                    log.info("Attachment %r not found for node %s", slug, node['_id'])
-                    link = ''
-                else:
-                    link = att.get('link', '')
-                    if link == 'self':
-                        link = " link='self'"
-                    elif link == 'custom':
-                        url = att.get('link_custom')
-                        if url:
-                            link = " link='%s'" % html.escape(url)
-                replaced = True
-                return '{attachment %r%s}' % (slug.replace(' ', '-'), link)
-
-            def update_markdown(value: str) -> str:
-                return old_slug_re.sub(replace, value)
-
-            iter_markdown(proj_node_types, node, update_markdown)
-
-            # Remove no longer used properties from attachments
-            new_attachments = {}
-            for slug, attachment in attachments.items():
-                replaced |= 'link' in attachment  # link_custom implies link
-                attachment.pop('link', None)
-                attachment.pop('link_custom', None)
-                new_attachments[slug.replace(' ', '-')] = attachment
-            node['properties']['attachments'] = new_attachments
-
-            if replaced:
-                total_nodes += 1
-            else:
-                # Nothing got replaced,
-                continue
-
-            if go:
-                # Use Eve to PUT, so we have schema checking.
-                db_node = remove_private_keys(node)
-                r, _, _, status = current_app.put_internal('nodes', db_node, _id=node['_id'])
-                if status != 200:
-                    log.error('Error %i storing altered node %s %s', status, node['_id'], r)
-                    failed_node_ids.add(node['_id'])
-                    # raise SystemExit('Error storing node; see log.')
-                log.debug('Updated node %s: %s', node['_id'], r)
-
-        log.info('Project %s (%s) has %d nodes with attachments',
-                 proj_url, proj_id, node_count)
-    log.info('%s %d nodes', 'Updated' if go else 'Would update', total_nodes)
-    if failed_node_ids:
-        log.warning('Failed to update %d of %d nodes: %s', len(failed_node_ids), total_nodes,
-                    ', '.join(str(nid) for nid in failed_node_ids))
-
-
 def _db_projects(proj_url: str, all_projects: bool, project_id='', *, go: bool) \
         -> typing.Iterable[dict]:
     """Yields a subset of the projects in the database.
@@ -1374,3 +1267,69 @@ def fix_projects_for_files(filepath: Path, go=False):
 
     log.info('Done updating %d files (found %d, modified %d) on %d projects',
              len(mapping), total_matched, total_modified, len(project_to_file_ids))
+
+
+@manager_maintenance.option('-u', '--user', dest='user', nargs='?',
+                            help='Update subscriptions for single user.')
+@manager_maintenance.option('-o', '--object', dest='context_object', nargs='?',
+                            help='Update subscriptions for context_object.')
+@manager_maintenance.option('-g', '--go', dest='go', action='store_true', default=False,
+                            help='Actually perform the changes (otherwise just show as dry-run).')
+def fix_missing_activities_subscription_defaults(user=None, context_object=None, go=False):
+    """Assign default values to activities-subscriptions documents where values are missing.
+    """
+
+    subscriptions_collection = current_app.db('activities-subscriptions')
+    lookup_is_subscribed = {
+        'is_subscribed': {'$exists': False},
+    }
+
+    lookup_notifications = {
+        'notifications.web': {'$exists': False},
+    }
+
+    if user:
+        lookup_is_subscribed['user'] = ObjectId(user)
+        lookup_notifications['user'] = ObjectId(user)
+
+    if context_object:
+        lookup_is_subscribed['context_object'] = ObjectId(context_object)
+        lookup_notifications['context_object'] = ObjectId(context_object)
+
+    num_need_is_subscribed_update = subscriptions_collection.count(lookup_is_subscribed)
+    log.info("Found %d documents that needs to be update 'is_subscribed'", num_need_is_subscribed_update)
+    num_need_notification_web_update = subscriptions_collection.count(lookup_notifications)
+    log.info("Found %d documents that needs to be update 'notifications.web'", num_need_notification_web_update)
+
+    if not go:
+        return
+
+    if num_need_is_subscribed_update > 0:
+        log.info("Updating 'is_subscribed'")
+        resp = subscriptions_collection.update(
+            lookup_is_subscribed,
+            {
+                '$set': {'is_subscribed': True}
+            },
+            multi=True,
+            upsert=False
+        )
+        if resp['nModified'] is not num_need_is_subscribed_update:
+            log.warning("Expected % documents to be update, was %d",
+                        num_need_is_subscribed_update, resp['nModified'])
+
+    if num_need_notification_web_update > 0:
+        log.info("Updating 'notifications.web'")
+        resp = subscriptions_collection.update(
+            lookup_notifications,
+            {
+                '$set': {'notifications.web': True}
+            },
+            multi=True,
+            upsert=False
+        )
+        if resp['nModified'] is not num_need_notification_web_update:
+            log.warning("Expected % documents to be update, was %d",
+                        num_need_notification_web_update, resp['nModified'])
+
+    log.info("Done updating 'activities-subscriptions' documents")
diff --git a/tests/test_api/test_cerberus.py b/tests/test_api/test_cerberus.py
index 01174f48..438c15fd 100644
--- a/tests/test_api/test_cerberus.py
+++ b/tests/test_api/test_cerberus.py
@@ -4,6 +4,8 @@ This'll help us upgrade to new versions of Cerberus.
 """
 
 import unittest
+
+from pillar.api.node_types.utils import markdown_fields
 from pillar.tests import AbstractPillarTest
 
 from bson import ObjectId
@@ -219,10 +221,9 @@ class MarkdownValidatorTest(AbstractSchemaValidationTest):
         'schema': {
             'type': 'dict',
             'schema': {
-                'content': {'type': 'string', 'required': True, 'validator': 'markdown'},
-                '_content_html': {'type': 'string'},
-                'descr': {'type': 'string', 'required': True, 'validator': 'markdown'},
-                '_descr_html': {'type': 'string'},
+                **markdown_fields('content', required=True),
+                **markdown_fields('descr', required=True),
+                'my_default_value': {'type': 'string', 'default': 'my default value'},
             }
         },
     }}
@@ -239,6 +240,7 @@ class MarkdownValidatorTest(AbstractSchemaValidationTest):
             '_content_html': '<h1>Header</h1>\n<p>Some text</p>\n',
             'descr': 'je moeder',
             '_descr_html': '<p>je moeder</p>\n',
+            'my_default_value': 'my default value'
         }]}
 
-        self.assertEqual(expect, doc)
+        self.assertEqual(expect, self.validator.document)
diff --git a/tests/test_api/test_notifications.py b/tests/test_api/test_notifications.py
new file mode 100644
index 00000000..66125891
--- /dev/null
+++ b/tests/test_api/test_notifications.py
@@ -0,0 +1,193 @@
+import bson
+import flask
+from bson import ObjectId
+
+from pillar.tests import AbstractPillarTest
+
+
+class NotificationsTest(AbstractPillarTest):
+    def setUp(self, **kwargs):
+        super().setUp(**kwargs)
+        self.project_id, _ = self.ensure_project_exists()
+        self.user1_id = self.create_user(user_id=str(bson.ObjectId()))
+        self.user2_id = self.create_user(user_id=str(bson.ObjectId()))
+
+    def test_create_node(self):
+        """When a node is created, a subscription should also be created"""
+
+        with self.app.app_context():
+            self.login_api_as(self.user1_id, roles={'subscriber', 'admin'},
+                              # This group is hardcoded in the EXAMPLE_PROJECT.
+                              group_ids=[ObjectId('5596e975ea893b269af85c0e')])
+
+            node_id = self.post_node(self.user1_id)
+
+            self.assertSubscribed(node_id, self.user1_id)
+            self.assertNotSubscribed(node_id, self.user2_id)
+
+    def test_comment_on_own_node(self):
+        """A comment on my own node should not give me any notifications"""
+
+        with self.app.app_context():
+            self.login_api_as(self.user1_id, roles={'subscriber', 'admin'},
+                              # This group is hardcoded in the EXAMPLE_PROJECT.
+                              group_ids=[ObjectId('5596e975ea893b269af85c0e')])
+
+            node_id = self.post_node(self.user1_id)
+            comment_id = self.post_comment(node_id)
+
+            self.assertSubscribed(comment_id, self.user1_id)
+            self.assertNotSubscribed(comment_id, self.user2_id)
+
+        with self.login_as(self.user1_id):
+            notification = self.notification_for_object(comment_id)
+            self.assertIsNone(notification)
+
+    def test_comment_on_node(self):
+        """A comment on some one else's node should give them a notification, and subscriptions should be created"""
+
+        with self.app.app_context():
+            self.login_api_as(self.user1_id, roles={'subscriber', 'admin'},
+                              # This group is hardcoded in the EXAMPLE_PROJECT.
+                              group_ids=[ObjectId('5596e975ea893b269af85c0e')])
+
+            node_id = self.post_node(self.user1_id)
+
+        with self.app.app_context():
+            self.login_api_as(self.user2_id, roles={'subscriber', 'admin'},
+                              # This group is hardcoded in the EXAMPLE_PROJECT.
+                              group_ids=[ObjectId('5596e975ea893b269af85c0e')])
+            comment_id = self.post_comment(node_id)
+
+            self.assertSubscribed(comment_id, self.user2_id)
+            self.assertNotSubscribed(comment_id, self.user1_id)
+            self.assertSubscribed(node_id, self.user1_id, self.user2_id)
+
+        with self.login_as(self.user1_id):
+            notification = self.notification_for_object(comment_id)
+            self.assertIsNotNone(notification)
+
+    def test_mark_notification_as_read(self):
+        with self.app.app_context():
+            self.login_api_as(self.user1_id, roles={'subscriber', 'admin'},
+                              # This group is hardcoded in the EXAMPLE_PROJECT.
+                              group_ids=[ObjectId('5596e975ea893b269af85c0e')])
+
+            node_id = self.post_node(self.user1_id)
+
+        with self.app.app_context():
+            self.login_api_as(self.user2_id, roles={'subscriber', 'admin'},
+                              # This group is hardcoded in the EXAMPLE_PROJECT.
+                              group_ids=[ObjectId('5596e975ea893b269af85c0e')])
+            comment_id = self.post_comment(node_id)
+
+        with self.login_as(self.user1_id):
+            notification = self.notification_for_object(comment_id)
+            self.assertFalse(notification['is_read'])
+
+            is_read_toggle_url = flask.url_for('notifications.action_read_toggle', notification_id=notification['_id'])
+            self.get(is_read_toggle_url)
+
+            notification = self.notification_for_object(comment_id)
+            self.assertTrue(notification['is_read'])
+
+            self.get(is_read_toggle_url)
+
+            notification = self.notification_for_object(comment_id)
+            self.assertFalse(notification['is_read'])
+
+    def test_unsubscribe(self):
+        """It should be possible to unsubscribe to notifications"""
+        with self.app.app_context():
+            self.login_api_as(self.user1_id, roles={'subscriber', 'admin'},
+                              # This group is hardcoded in the EXAMPLE_PROJECT.
+                              group_ids=[ObjectId('5596e975ea893b269af85c0e')])
+
+            node_id = self.post_node(self.user1_id)
+
+        with self.app.app_context():
+            self.login_api_as(self.user2_id, roles={'subscriber', 'admin'},
+                              # This group is hardcoded in the EXAMPLE_PROJECT.
+                              group_ids=[ObjectId('5596e975ea893b269af85c0e')])
+            comment_id = self.post_comment(node_id)
+
+        with self.login_as(self.user1_id):
+            notification = self.notification_for_object(comment_id)
+            self.assertTrue(notification['is_subscribed'])
+
+            is_subscribed_toggle_url =\
+                flask.url_for('notifications.action_subscription_toggle', notification_id=notification['_id'])
+            self.get(is_subscribed_toggle_url)
+
+            notification = self.notification_for_object(comment_id)
+            self.assertFalse(notification['is_subscribed'])
+
+        with self.app.app_context():
+            self.login_api_as(self.user2_id, roles={'subscriber', 'admin'},
+                              # This group is hardcoded in the EXAMPLE_PROJECT.
+                              group_ids=[ObjectId('5596e975ea893b269af85c0e')])
+            comment2_id = self.post_comment(node_id)
+
+        with self.login_as(self.user1_id):
+            notification = self.notification_for_object(comment2_id)
+            self.assertFalse(notification['is_subscribed'])
+
+    def assertSubscribed(self, node_id, *user_ids):
+        subscriptions_col = self.app.data.driver.db['activities-subscriptions']
+
+        lookup = {
+            'context_object': node_id,
+            'notifications.web': True,
+        }
+        subscriptions = list(subscriptions_col.find(lookup))
+        self.assertEquals(len(subscriptions), len(user_ids))
+        for s in subscriptions:
+            self.assertIn(s['user'], user_ids)
+            self.assertEquals(s['context_object_type'], 'node')
+
+    def assertNotSubscribed(self, node_id, user_id):
+        subscriptions_col = self.app.data.driver.db['activities-subscriptions']
+
+        lookup = {
+            'context_object': node_id,
+        }
+        subscriptions = subscriptions_col.find(lookup)
+        for s in subscriptions:
+            self.assertNotEquals(s['user'], user_id)
+
+    def notification_for_object(self, node_id):
+        notifications_url = flask.url_for('notifications.index')
+        notification_items = self.get(notifications_url).json['items']
+
+        object_url = flask.url_for('nodes.redirect_to_context', node_id=str(node_id), _external=False)
+        for candidate in notification_items:
+            if candidate['object_url'] == object_url:
+                return candidate
+
+    def post_node(self, user_id):
+        node_doc = {'description': '',
+                    'project': self.project_id,
+                    'node_type': 'asset',
+                    'user': user_id,
+                    'properties': {'status': 'published',
+                                   'tags': [],
+                                   'order': 0,
+                                   'categories': '',
+                                   },
+                    'name': 'My first test node'}
+
+        r, _, _, status = self.app.post_internal('nodes', node_doc)
+        self.assertEqual(status, 201, r)
+        node_id = r['_id']
+        return node_id
+
+    def post_comment(self, node_id):
+        comment_url = flask.url_for('nodes_api.post_node_comment', node_path=str(node_id))
+        comment = self.post(
+            comment_url,
+            json={
+                'msg': 'je möder lives at [home](https://cloud.blender.org/)',
+            },
+            expected_status=201,
+        )
+        return ObjectId(comment.json['id'])
diff --git a/tests/test_cli/test_maintenance.py b/tests/test_cli/test_maintenance.py
index b268c837..1710ec3d 100644
--- a/tests/test_cli/test_maintenance.py
+++ b/tests/test_cli/test_maintenance.py
@@ -114,147 +114,6 @@ class UpgradeAttachmentSchemaTest(AbstractPillarTest):
             self.assertEqual({}, db_node_type['form_schema'])
 
 
-class UpgradeAttachmentUsageTest(AbstractPillarTest):
-    def setUp(self, **kwargs):
-        super().setUp(**kwargs)
-        self.pid, self.uid = self.create_project_with_admin(user_id=24 * 'a')
-
-        with self.app.app_context():
-            files_coll = self.app.db('files')
-
-            res = files_coll.insert_one({
-                **ctd.EXAMPLE_FILE,
-                'project': self.pid,
-                'user': self.uid,
-            })
-            self.fid = res.inserted_id
-
-    def test_image_link(self):
-        with self.app.app_context():
-            nodes_coll = self.app.db('nodes')
-            res = nodes_coll.insert_one({
-                **ctd.EXAMPLE_NODE,
-                'picture': self.fid,
-                'project': self.pid,
-                'user': self.uid,
-                'description': "# Title\n\n@[slug 0]\n@[slug1]\n@[slug2]\nEitje van Fabergé.",
-                'properties': {
-                    'status': 'published',
-                    'content_type': 'image',
-                    'file': self.fid,
-                    'attachments': {
-                        'slug 0': {
-                            'oid': self.fid,
-                            'link': 'self',
-                        },
-                        'slug1': {
-                            'oid': self.fid,
-                            'link': 'custom',
-                            'link_custom': 'https://cloud.blender.org/',
-                        },
-                        'slug2': {
-                            'oid': self.fid,
-                        },
-                    }
-                }
-            })
-            nid = res.inserted_id
-
-        from pillar.cli.maintenance import upgrade_attachment_usage
-
-        with self.app.app_context():
-            upgrade_attachment_usage(proj_url=ctd.EXAMPLE_PROJECT['url'], go=True)
-            node = nodes_coll.find_one({'_id': nid})
-
-            self.assertEqual(
-                "# Title\n\n"
-                "{attachment 'slug-0' link='self'}\n"
-                "{attachment 'slug1' link='https://cloud.blender.org/'}\n"
-                "{attachment 'slug2'}\n"
-                "Eitje van Fabergé.",
-                node['description'],
-                'The description should be updated')
-            self.assertEqual(
-                "<h1>Title</h1>\n"
-                "<!-- {attachment 'slug-0' link='self'} -->\n"
-                "<!-- {attachment 'slug1' link='https://cloud.blender.org/'} -->\n"
-                "<!-- {attachment 'slug2'} -->\n"
-                "<p>Eitje van Fabergé.</p>\n",
-                node['_description_html'],
-                'The _description_html should be updated')
-
-            self.assertEqual(
-                {'slug-0': {'oid': self.fid},
-                 'slug1': {'oid': self.fid},
-                 'slug2': {'oid': self.fid},
-                 },
-                node['properties']['attachments'],
-                'The link should have been removed from the attachment')
-
-    def test_post(self):
-        """This requires checking the dynamic schema of the node."""
-        with self.app.app_context():
-            nodes_coll = self.app.db('nodes')
-            res = nodes_coll.insert_one({
-                **ctd.EXAMPLE_NODE,
-                'node_type': 'post',
-                'project': self.pid,
-                'user': self.uid,
-                'picture': self.fid,
-                'description': "meh",
-                'properties': {
-                    'status': 'published',
-                    'content': "# Title\n\n@[slug0]\n@[slug1]\n@[slug2]\nEitje van Fabergé.",
-                    'attachments': {
-                        'slug0': {
-                            'oid': self.fid,
-                            'link': 'self',
-                        },
-                        'slug1': {
-                            'oid': self.fid,
-                            'link': 'custom',
-                            'link_custom': 'https://cloud.blender.org/',
-                        },
-                        'slug2': {
-                            'oid': self.fid,
-                        },
-                    }
-                }
-            })
-            nid = res.inserted_id
-
-        from pillar.cli.maintenance import upgrade_attachment_usage
-
-        with self.app.app_context():
-            upgrade_attachment_usage(proj_url=ctd.EXAMPLE_PROJECT['url'], go=True)
-            node = nodes_coll.find_one({'_id': nid})
-
-            self.assertEqual(
-                "# Title\n\n"
-                "{attachment 'slug0' link='self'}\n"
-                "{attachment 'slug1' link='https://cloud.blender.org/'}\n"
-                "{attachment 'slug2'}\n"
-                "Eitje van Fabergé.",
-                node['properties']['content'],
-                'The content should be updated')
-            self.assertEqual(
-                "<h1>Title</h1>\n"
-                "<!-- {attachment 'slug0' link='self'} -->\n"
-                "<!-- {attachment 'slug1' link='https://cloud.blender.org/'} -->\n"
-                "<!-- {attachment 'slug2'} -->\n"
-                "<p>Eitje van Fabergé.</p>\n",
-                node['properties']['_content_html'],
-                'The _content_html should be updated')
-
-            self.assertEqual(
-                {'slug0': {'oid': self.fid},
-                 'slug1': {'oid': self.fid},
-                 'slug2': {'oid': self.fid},
-                 },
-                node['properties']['attachments'],
-                'The link should have been removed from the attachment')
-
-
 class ReconcileNodeDurationTest(AbstractPillarTest):
     def setUp(self, **kwargs):
         super().setUp(**kwargs)
@@ -459,3 +318,52 @@ class DeleteProjectlessFilesTest(AbstractPillarTest):
             self.assertEqual(file1_doc, found1)
             self.assertEqual(file3_doc, found3)
             self.assertEqual(file3_doc, found3)
+
+
+class FixMissingActivitiesSubscription(AbstractPillarTest):
+    def test_fix_missing_activities_subscription(self):
+        from pillar.cli.maintenance import fix_missing_activities_subscription_defaults
+
+        with self.app.app_context():
+            subscriptions_collection = self.app.db('activities-subscriptions')
+
+            invalid_subscription = {
+                'user': ObjectId(),
+                'context_object_type': 'node',
+                'context_object': ObjectId(),
+            }
+
+            valid_subscription = {
+                'user': ObjectId(),
+                'context_object_type': 'node',
+                'context_object': ObjectId(),
+                'is_subscribed': False,
+                'notifications': {'web': False, }
+            }
+
+            result = subscriptions_collection.insert_one(
+                invalid_subscription,
+                bypass_document_validation=True)
+
+            id_invalid = result.inserted_id
+
+            result = subscriptions_collection.insert_one(
+                valid_subscription,
+                bypass_document_validation=True)
+
+            id_valid = result.inserted_id
+
+            fix_missing_activities_subscription_defaults()  # Dry run. Nothing should change
+            invalid_subscription = subscriptions_collection.find_one({'_id': id_invalid})
+            self.assertNotIn('is_subscribed', invalid_subscription.keys())
+            self.assertNotIn('notifications', invalid_subscription.keys())
+
+            fix_missing_activities_subscription_defaults(go=True)  # Real run. Invalid should be updated
+            invalid_subscription = subscriptions_collection.find_one({'_id': id_invalid})
+            self.assertTrue(invalid_subscription['is_subscribed'])
+            self.assertTrue(invalid_subscription['notifications']['web'])
+
+            # Was already ok. Should not have been updated
+            valid_subscription = subscriptions_collection.find_one({'_id': id_valid})
+            self.assertFalse(valid_subscription['is_subscribed'])
+            self.assertFalse(valid_subscription['notifications']['web'])