feat: make the capsule subscribable over Gemini

Link labels in every capsule index are now ISO 8601 and lead with the date,
which is what Gemini's subscription convention reads a feed entry from: the
first ten characters of a label must parse as YYYY-MM-DD or the line is
ignored. The board's `09 Oct 26` parsed as nothing, so the front page, the
archive and all six areas registered zero entries and the capsule was
invisible to every aggregator in Geminispace. `/archive` now carries 36.

The Connect page advertised `/feeds/all.rss.xml` and `/feeds/all.atom.xml`,
which exist only in the HTML build and 404 over Gemini. That paragraph now
sits behind an itsybitsy `only html` format gate, with prose outside it that
is true on both faces. The gate can only cut that way round: Pelican does not
implement the directive and renders it as an inert comment, so a gate
subtracts from the capsule and never from the site.

Also adds dormant `SMOLNET_COMMENTS` support, for a per-article thread link
gated to the text formats. Off until kapara is deployed, since setting it
would put a dead link under every article.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

plugins/smolnet.py

@@ -25,6 +25,8 @@ setting, and every setting has a default, so a site can switch it on with just
     SMOLNET_SIDECAR       '.itsybitsy.toml'
     SMOLNET_ROT13_ATTR    None        attribute holding a rot13 address to reveal
     SMOLNET_TAGLINE       None        one line at the foot of the index
+    SMOLNET_COMMENTS      None        thread URL per article, `{slug}` substituted
+    SMOLNET_COMMENTS_FORMATS          formats that get it, default gemtext and text
 
 Why link-only paragraphs rather than lists: gemtext has one kind of link, a line
 beginning `=>`, so a renderer has to lift links out of the prose. itsybitsy's
@@ -58,6 +60,7 @@ LABELS = {
     "archive": "All entries",
     "pages": "Pages",
     "back": "Back to the index",
+    "comments": "Comments",
 }
 
 # Pelican's metadata block: bare `Key: value` lines from the first line on,
@@ -240,6 +243,22 @@ def link(label, href):
     return "[%s](%s)" % (label.replace("]", ""), href)
 
 
+def gate(block, formats):
+    """Wrap a block in an itsybitsy format gate.
+
+    A gate names output format ids, not protocols: `gemtext` is the one Gemini,
+    Spartan, Nex and Gopher are all served from, so there is no gating on a
+    protocol name. Blank lines because a gate encloses *blocks*, and a
+    directive run together with the line under it is one block.
+
+    Note which way this can cut. Pelican does not implement the directive and
+    renders it as an inert HTML comment, so a gate can only ever subtract from
+    the capsule, never from the site. That is the right direction for anything
+    this plugin writes, because the capsule is all it writes.
+    """
+    return "<!-- only %s -->\n\n%s\n\n<!-- end -->" % (" ".join(formats), block)
+
+
 class CapsuleGenerator(Generator):
     """Writes the capsule once the content generators have read everything."""
 
@@ -257,6 +276,14 @@ class CapsuleGenerator(Generator):
         dead_hosts = {h.lower().lstrip(".") for h in
                       s.get("LOSTIMAGES_DEAD_HOSTS", [])}
         sidecar_name = s.get("SMOLNET_SIDECAR", ".itsybitsy.toml")
+        # kapara serves the same remark42 thread the HTML widget shows, over
+        # Gemini, Spartan and Nex. Articles only: the widget is on articles
+        # only, and a page with a thread nobody can reach from the site is a
+        # second conversation, which is the one thing the gateway exists to
+        # avoid. Gated because this tree can also be served as WML or
+        # XHTML-MP, and a handset cannot open a gemini:// link.
+        comments = s.get("SMOLNET_COMMENTS")
+        comments_formats = s.get("SMOLNET_COMMENTS_FORMATS", ("gemtext", "text"))
 
         articles = [a for a in self.context.get("articles", [])]
         pages = [p for p in self.context.get("pages", [])]
@@ -274,7 +301,7 @@ class CapsuleGenerator(Generator):
         os.makedirs(out, exist_ok=True)
         root_titles = {}
 
-        def emit(doc, trail):
+        def emit(doc, trail, thread=False):
             body = strip_metadata(self._read(doc))
             body = body.replace(SUMMARY_MARKER, "")
             body = rewrite_internal(body, by_source, self._warn)
@@ -288,6 +315,11 @@ class CapsuleGenerator(Generator):
             if trail:
                 head += "%s\n\n" % trail
             text = head + body.strip() + "\n"
+            if thread and comments:
+                text += "\n%s\n" % gate(
+                    link(labels["comments"], comments.format(slug=doc.slug)),
+                    comments_formats,
+                )
             name = "%s.md" % doc.slug
             with open(os.path.join(out, name), "w", encoding="utf-8") as fh:
                 fh.write(text)
@@ -308,7 +340,7 @@ class CapsuleGenerator(Generator):
                     bits.append(value.strftime("%d %b %Y"))
                 else:
                     bits.append(str(value))
-            emit(article, " · ".join(bits))
+            emit(article, " · ".join(bits), thread=True)
         for page in pages:
             emit(page, "")
 
@@ -381,12 +413,25 @@ class CapsuleGenerator(Generator):
         relative targets would need `../` and would break again at any further
         depth; `/slug` is correct from anywhere and the daemon resolves it
         against the site root.
+
+        The date is ISO 8601 and it leads the label, which is protocol surface
+        rather than a style choice. Gemini's subscription convention takes a
+        link line for a feed entry when the first ten characters of its label
+        are a `YYYY-MM-DD` date, reads the rest as the title, and ignores every
+        line where they are not, so this one format is what makes each index a
+        feed: the front page, the archive and all six areas, subscribable in a
+        client and submittable to an aggregator. A board's `09 Oct 26` parses
+        as nothing and registers zero entries.
+
+        The separator is the one the convention's own examples use. Clients are
+        asked to sanitise whatever sits between the date and the title, but not
+        all of them do, and a `·` that survives arrives as part of the title.
         """
         rows = []
         for doc in docs:
             label = doc.title
             if getattr(doc, "date", None):
-                label = "%s · %s" % (doc.date.strftime("%d %b %y"), doc.title)
+                label = "%s - %s" % (doc.date.strftime("%Y-%m-%d"), doc.title)
             rows.append(link(label, "/%s" % doc.slug))
         # Hard breaks, which is two trailing spaces in CommonMark. gemtext lifts
         # each link to its own => line either way, but the inline formats lay a

randogoth.com/content/pages/connect.md

@@ -19,5 +19,10 @@ Slug: connect
 
 ## Feeds
 
+<!-- only html -->
 This site has [RSS](/feeds/all.rss.xml) and [Atom](/feeds/all.atom.xml), and so
-does every area. Readers are better than algorithms.
+does every area.
+<!-- end -->
+Over Gemini there is no XML and none is needed: every listing is ISO dated, so a
+client can subscribe to the main menu, the archive or a single area as it
+stands. Readers are better than algorithms.

randogoth.com/pelicanconf.py

@@ -37,12 +37,27 @@ SMOLNET_SECTION_ORDER = BBS_AREAS
 SMOLNET_LABELS = {
     'sections': 'Message areas', 'latest': 'Latest messages',
     'archive': 'All messages', 'back': 'Main menu', 'pages': 'System',
+    'comments': 'Replies',
 }
 # The address is rot13 in the HTML because of web crawlers. That defence needs
 # JavaScript, which Gemini does not have, and the renderer drops raw HTML, so
 # without this the one thing the Connect page exists to carry would vanish.
 SMOLNET_ROT13_ATTR = 'data-r'
 
+# The capsule's half of the replies box, off until kapara is deployed. Setting
+# it puts a Replies link under every article in the capsule, so it must not be
+# switched on before something answers Gemini on the other end.
+#
+# When it lands: kapara gateways this same remark42 instance, and remark42 keys
+# a thread by the article's canonical web URL, which bbs.js builds as origin +
+# pathname. So kapara has to run with `-page-url 'https://randogoth.com/%s.html'`
+# and `-site randogoth`, or the capsule opens a second, empty thread under every
+# article instead of joining the one that is there. Port 1966, not the default:
+# stunnel owns 1965 in front of itsybitsy, and kapara has to terminate its own
+# TLS because Gemini client certificates are its identity mechanism.
+#
+# SMOLNET_COMMENTS = 'gemini://comments.randogoth.com:1966/c/{slug}'
+
 # Articles live flat in content/; assets are carved out of that.
 ARTICLE_PATHS = ['']
 ARTICLE_EXCLUDES = ['pages', 'library', 'images', 'files']