@geekity/cms 0.3.0 → 0.5.0

package/dist/web/routes.js

@@ -1,14 +1,25 @@
+import { listUsers } from '../admin/accounts.js';
+import { readSiteSettings } from '../admin/settings.js';
 import { serializeDocument } from '../content/writer.js';
-import { assetNotModified, assetResponse, findThemeAsset, findUpload, matchesEtag, themeAssetNotModified, themeAssetResponse, themeSearchPath, THEME_ASSET_PREFIX, UPLOAD_ASSET_MAX_AGE, UPLOAD_ASSET_PREFIX, } from './assets.js';
+import { findImageVariant, VARIANT_ASSET_PREFIX } from '../images/variants.js';
+import { assetNotModified, assetResponse, findThemeAsset, findUpload, matchesEtag, themeAssetNotModified, themeAssetResponse, THEME_ASSET_PREFIX, UPLOAD_ASSET_MAX_AGE, UPLOAD_ASSET_PREFIX, } from './assets.js';
+import { COMMENT_NOTICE_PARAM, COMMENT_REPLY_PARAM } from '../comments/form.js';
+import { commentNoticeFor, commentReplyTarget, mountComments } from '../comments/routes.js';
+import { signedInCommenter } from '../comments/viewer.js';
+import { CONTACT_NOTICE_PARAM, contactNoticeFor } from '../contact/form.js';
+import { mountContact } from '../contact/routes.js';
+import { mountWebmentions, WEBMENTION_PATH } from '../webmention/routes.js';
+import { mountNotificationLinks } from '../notifications/routes.js';
+import { authorFeedHref, authorHref, authorNames, parseAuthorPath, profileContext, } from './authors.js';
+import { feedComments, spokenIn } from './conversation.js';
 import { isPublicDocument, publicDocumentAt } from './documents.js';
-import { feedResponse, feedSize, FEED_FILES } from './feeds.js';
-import { DOCUMENT_REPRESENTATIONS, documentJson, LISTING_REPRESENTATIONS, lastModifiedOf, latestModified, notAcceptableResponse, representationEtag, representationResponse, selectRepresentation, splitRepresentationExtension, } from './negotiate.js';
+import { commentsFeedPath, commentsFeedResponse, feedPathUnder, feedResponse, feedSize, splitFeedPath, COMMENTS_ROOT, COMMENTS_TITLE_PREFIX, FEED_FORMATS, FEED_SEGMENTS, } from './feeds.js';
+import { DOCUMENT_REPRESENTATIONS, documentJson, LISTING_REPRESENTATIONS, lastModifiedOf, latestModified, MEDIA_TYPES, notAcceptableResponse, representationHref, representationEtag, representationResponse, selectRepresentation, splitRepresentationExtension, } from './negotiate.js';
 import { offsetForPage, paginate } from './pagination.js';
 import { TEMPLATES } from './render.js';
-/** Where a paginated listing's later pages live, under any listing root. */
-export const PAGE_SEGMENT = 'page';
-/** Root of the tag archives. */
-export const TAG_SEGMENT = 'tags';
+import { robotsResponse, sitemapResponse, ROBOTS_PATH, SITEMAP_CHILD_ROUTE, SITEMAP_PATH, } from './sitemap.js';
+import { searchHref, searchJson, searchPageIndex, searchQuery, SEARCH_PAGE_PARAM, SEARCH_PATH, SEARCH_QUERY_PARAM, } from './search.js';
+import { PAGE_SEGMENT, redirectedTerm, taxonomyForSegment, termHref } from './taxonomy.js';
 /**
  * Register the public site on a Hono app.
  *
@@ -19,84 +30,365 @@ export const TAG_SEGMENT = 'tags';
  */
 export function mountPublicSite(app) {
     app.get(`${THEME_ASSET_PREFIX}*`, themeAsset);
+    // Derived images go on first: their prefix is inside the uploads one, so the
+    // general route would otherwise swallow them and answer 404 for a file that
+    // is not under `content/uploads/` at all.
+    app.get(`${VARIANT_ASSET_PREFIX}*`, imageVariant);
     app.get(`${UPLOAD_ASSET_PREFIX}*`, upload);
-    app.get('/', (c) => listing(c, { tag: undefined, pageNumber: 0 }));
+    // Where the comment form under a post posts to (TASK-50). A POST at a fixed
+    // path of the CMS's own, so no permalink can ever shadow it and the route
+    // table does not grow with the site.
+    mountComments(app);
+    // And where a contact form's message is sent (TASK-56): the same prefix
+    // again, so the page a form is on can be permalinked anywhere.
+    mountContact(app);
+    // And where a webmention is sent (TASK-51), for the same reason and under
+    // the same prefix.
+    mountWebmentions(app);
+    // And where the one-click links in a notification land (TASK-55): the same
+    // prefix again, and no session behind either of them.
+    mountNotificationLinks(app);
+    app.get('/', (c) => {
+        // The site's latest posts, or the page the Reading setting names — the
+        // whole of WordPress's "Your homepage displays", decided per request off
+        // `site.json` so a save on the settings screen moves the front page on the
+        // very next one.
+        const home = frontPages(c).home;
+        if (home === undefined)
+            return listing(c, { term: undefined, pageNumber: 0 });
+        return negotiateDocument(c, home, selectFromAccept(c, DOCUMENT_REPRESENTATIONS), '/');
+    });
     app.get(`/${PAGE_SEGMENT}/:page{[0-9]+}/`, (c) => {
         const requested = Number(c.req.param('page'));
         // Page one is the home page; it does not get a second URL.
         if (requested <= 1)
             return c.redirect('/', 301);
-        return listing(c, { tag: undefined, pageNumber: requested - 1 });
-    });
-    // Feeds are routes rather than representations, so a reader's careless
-    // `Accept` header cannot land it on the HTML (doc-3). They are registered
-    // before the tag archive so the shape of the URL, not the negotiator,
-    // decides what comes back.
-    app.get(`/${FEED_FILES.atom}`, (c) => feed(c, 'atom', undefined));
-    app.get(`/${FEED_FILES.json}`, (c) => feed(c, 'json', undefined));
-    app.get(`/${TAG_SEGMENT}/:tag/${FEED_FILES.atom}`, (c) => feed(c, 'atom', c.req.param('tag')));
-    app.get(`/${TAG_SEGMENT}/:tag/${FEED_FILES.json}`, (c) => feed(c, 'json', c.req.param('tag')));
-    app.get(`/${TAG_SEGMENT}/:tag/`, (c) => listing(c, { tag: c.req.param('tag'), pageNumber: 0 }));
-    app.get(`/${TAG_SEGMENT}/:tag/${PAGE_SEGMENT}/:page{[0-9]+}/`, (c) => {
-        const tag = c.req.param('tag');
-        const requested = Number(c.req.param('page'));
-        if (requested <= 1)
-            return c.redirect(tagHref(tag, 0), 301);
-        return listing(c, { tag, pageNumber: requested - 1 });
+        const pages = frontPages(c);
+        // With a posts page, the listing's pages live under it and these are the
+        // URLs they used to have; with a static homepage and no posts page, the
+        // listing has no page of its own and neither have its pages.
+        if (pages.posts !== undefined) {
+            return c.redirect(listingPageHref(pages.posts.permalink, requested - 1), 301);
+        }
+        if (pages.home !== undefined)
+            return notFound(c);
+        return listing(c, { term: undefined, pageNumber: requested - 1 });
     });
-    app.notFound(resolveDocument);
+    // The site's own feeds are routes rather than representations, so a reader's
+    // careless `Accept` header cannot land it on the HTML (doc-3), and so a
+    // document permalinked at `/feed/` cannot take the subscribers' URL. Their
+    // paths are fixed — only the taxonomy feeds hang off a configurable base —
+    // so the route table can hold them.
+    for (const format of FEED_FORMATS) {
+        app.get(feedPathUnder('/', format), (c) => feed(c, format, {}));
+    }
+    // The site-wide comments feed, for the same reason and at WordPress's URL.
+    app.get(commentsFeedHref(undefined), (c) => comments(c, undefined));
+    // The sitemap, its children and the robots file: fixed paths at the root of
+    // the site, which is the only place a crawler looks, and routes for the same
+    // reason the feeds are — no document permalinked there can take the URL a
+    // search engine polls.
+    app.get(SITEMAP_PATH, (c) => sitemap(c, undefined));
+    app.get(SITEMAP_CHILD_ROUTE, (c) => sitemap(c, Number(c.req.param('page'))));
+    app.get(ROBOTS_PATH, (c) => robotsResponse(c.var.config.baseUrl, conditionalHeaders(c)));
+    // The site's search (TASK-22): a route at a fixed path for the reason the
+    // feeds are, so the form in every theme's footer submits somewhere no
+    // permalink can take. It negotiates HTML and JSON like a listing, and
+    // `/search/index.json` is the same escape hatch a listing has. The path
+    // without its slash is where a hand-typed URL lands, and keeps its query on
+    // the way to the real one.
+    app.get(SEARCH_PATH, (c) => search(c, selectFromAccept(c, LISTING_REPRESENTATIONS)));
+    app.get(representationHref(SEARCH_PATH, 'json'), (c) => search(c, 'json'));
+    app.get(SEARCH_PATH.slice(0, -1), (c) => c.redirect(`${SEARCH_PATH}${new URL(c.req.url).search}`, 301));
+    // The taxonomy archives are deliberately not routes. A route table is fixed
+    // when the app is built and the bases are a setting, so an archive is
+    // resolved per request in the not-found handler, from the base the site
+    // holds at that moment (TASK-36).
+    app.notFound(resolveRequest);
 }
 /**
- * The last stop for a request: the document at this URL, the same document in
- * the representation a `.md` or `.json` suffix asked for, the canonical URL it
- * should have asked for, or the theme's 404.
+ * The last stop for a request: a taxonomy archive or one of its feeds, the
+ * document at this URL, the same document in the representation a `.md` or
+ * `.json` suffix asked for, the canonical URL it should have asked for, or the
+ * theme's 404.
  *
- * The suffix is stripped and the *same* {@link publicDocumentAt} lookup runs
- * again, so no representation can resolve to a document the others cannot see.
+ * The archives come first, so an archive URL means the archive however a
+ * document is permalinked; they are here rather than in the route table
+ * because their bases are a setting, and a route table is fixed when the app
+ * is built. The suffix is stripped and the *same* {@link publicDocumentAt}
+ * lookup runs again, so no representation can resolve to a document the others
+ * cannot see.
  */
-function resolveDocument(c) {
-    const { store } = c.var;
+function resolveRequest(c) {
+    const { store, renderer } = c.var;
     const pathname = requestPath(c);
+    const bases = renderer.taxonomyBases();
+    const authors = authorsOf(c);
+    const pages = frontPages(c);
+    // The posts page carries the listing at its own URL, so it is answered
+    // before the document lookup that would otherwise render it as the page it
+    // also is. Its pagination hangs off it: `{permalink}page/2/`.
+    if (pages.posts !== undefined) {
+        const request = postsPageRequest(pages.posts, pathname);
+        if (request !== undefined) {
+            // `{permalink}page/1/` is the listing's own URL spelled twice, and
+            // collapses onto it exactly as `/page/1/` collapses onto `/`.
+            if (request.pageNumber === 0 && pathname !== pages.posts.permalink) {
+                return c.redirect(encodePath(pages.posts.permalink), 301);
+            }
+            return listing(c, request);
+        }
+    }
+    const feedRequest = parseFeedPath(store, pathname, bases, authors);
+    if (feedRequest !== undefined) {
+        const { target, format } = feedRequest;
+        if (!feedRequest.canonical)
+            return c.redirect(feedTargetHref(target, format, bases), 301);
+        return target.kind === 'listing'
+            ? feed(c, format, target.subject)
+            : comments(c, target.document);
+    }
+    const archive = taxonomyArchive(c, pathname, bases, authors);
+    if (archive !== undefined)
+        return archive;
+    // And one person's, which is the same kind of thing at a path the site
+    // reserves: `author` is never a document's first segment, so this can be
+    // answered before the permalink lookup without anything being shadowed.
+    const byAuthor = authorArchive(c, pathname, bases, authors);
+    if (byAuthor !== undefined)
+        return byAuthor;
     const document = publicDocumentAt(store, pathname);
     if (document !== undefined) {
+        // The front page has one URL. The page it is made of keeps its own
+        // permalink in the file, so that a homepage put back to being an ordinary
+        // page goes back to answering there, and that URL points at `/` while it
+        // is the front page.
+        if (document.path === pages.home?.path)
+            return c.redirect('/', 301);
+        // WordPress answered `?feed=rss2` on a permalink with that post's comments
+        // feed, which is the only feed a post has here too. The other spellings
+        // name a format a comments feed does not come in, so they are not a feed
+        // request at all and the page is served.
+        if (document.type === 'post' && queryFeedFormat(c) === 'rss') {
+            return c.redirect(commentsFeedHref(document), 301);
+        }
         return negotiateDocument(c, document, selectFromAccept(c, DOCUMENT_REPRESENTATIONS));
     }
     const extension = splitRepresentationExtension(pathname);
     if (extension !== undefined) {
         for (const candidate of extension.paths) {
-            const found = publicDocumentAt(store, candidate);
-            if (found !== undefined)
-                return negotiateDocument(c, found, extension.representation);
+            // `/index.md` and `/index.json` are the front page's own
+            // representations on a site whose `/` is a page: the document is at `/`
+            // however it is asked for.
+            const found = candidate === '/' ? pages.home : publicDocumentAt(store, candidate);
+            if (found !== undefined) {
+                return negotiateDocument(c, found, extension.representation, candidate === '/' ? '/' : undefined);
+            }
         }
-        // `/index.json`, `/page/2/index.json`, `/tags/x/index.json`: the same
-        // escape hatch over a listing, for the representations a listing has.
+        // `/index.json`, `/page/2/index.json`, `/tag/x/index.json`,
+        // `/category/x/index.json`: the same escape hatch over a listing, for the
+        // representations a listing has.
         if (LISTING_REPRESENTATIONS.includes(extension.representation)) {
             for (const candidate of extension.paths) {
-                const request = parseListingPath(candidate);
+                const request = listingRequestAt(pages, candidate, bases, authors);
                 if (request !== undefined)
                     return listing(c, request, extension.representation);
             }
         }
     }
-    const canonical = canonicalPath(c, pathname);
+    const canonical = canonicalPath(c, pathname, bases, authors);
     if (canonical !== undefined)
         return c.redirect(canonical, 301);
     return notFound(c);
 }
+/**
+ * One taxonomy archive, the redirect that puts it at its canonical URL, or
+ * `undefined` when the path is not one.
+ *
+ * `/{base}/x/page/1/` collapses onto the archive root the way `/page/1/`
+ * collapses onto the home page, and an archive nothing carries 404s rather
+ * than redirecting first: a URL that leads nowhere should cost one 404, not a
+ * redirect and then a 404.
+ */
+function taxonomyArchive(c, pathname, bases, authors) {
+    if (!pathname.endsWith('/'))
+        return undefined;
+    const request = parseListingPath(pathname, bases, authors);
+    if (request?.term === undefined)
+        return undefined;
+    // An archive that answers is served; only once nothing carries the term is
+    // the record of renames consulted, so a term that comes back into use — or
+    // one that was merged into and then recreated — beats what it used to be
+    // called. One hop: the list is stored with its chains already collapsed.
+    if (countListing(c.var.store, request) === 0) {
+        const moved = movedTerm(c, request.term);
+        if (moved !== undefined)
+            return c.redirect(termHref(moved, request.pageNumber, bases), 301);
+    }
+    const canonical = termHref(request.term, request.pageNumber, bases);
+    if (canonical !== encodePath(pathname)) {
+        return countListing(c.var.store, request) === 0 ? notFound(c) : c.redirect(canonical, 301);
+    }
+    return listing(c, request);
+}
+/**
+ * One person's archive, the redirect that puts it at its canonical URL, or
+ * `undefined` when the path is not one.
+ *
+ * Unlike a taxonomy archive it does not depend on there being anything on it.
+ * A tag exists only while something carries it; a person exists because they
+ * have an account, and after decision-14 this URL is the id their followers
+ * file them under — so an archive that 404'd until its owner published would
+ * be an account that came into being with a post. What decides the 404 is
+ * whether the username is somebody's, which {@link parseListingPath} has
+ * already asked.
+ *
+ * `/author/{username}/page/1/` collapses onto the archive root the way
+ * `/page/1/` collapses onto the home page.
+ */
+function authorArchive(c, pathname, bases, authors) {
+    if (!pathname.endsWith('/'))
+        return undefined;
+    const request = parseListingPath(pathname, bases, authors);
+    if (request?.author === undefined)
+        return undefined;
+    const canonical = authorHref(request.author.user.username, request.pageNumber);
+    if (canonical !== encodePath(pathname))
+        return c.redirect(canonical, 301);
+    return listing(c, request);
+}
+/**
+ * Where a term with nothing left under it went, or `undefined` when the site
+ * records no such rename.
+ */
+function movedTerm(c, term) {
+    const moved = redirectedTerm(c.var.renderer.termRedirects(), term);
+    return moved === undefined ? undefined : { taxonomy: term.taxonomy, term: moved };
+}
+/**
+ * A path as the feed it names: `/feed/`, `/feed/atom/`, `/{base}/x/feed/json/`,
+ * the comments feeds at `/comments/feed/` and `{permalink}feed/`, and the
+ * `/feed/rss/` spelling WordPress also answered at any of those roots.
+ *
+ * The site's own canonical feeds are real routes and never reach here; what
+ * does reach here is every taxonomy feed — their bases are a setting, so they
+ * cannot be in the route table — every post's comments feed, and every alias.
+ */
+function parseFeedPath(store, pathname, bases, authors) {
+    const split = splitFeedPath(pathname);
+    if (split === undefined)
+        return undefined;
+    const { format, canonical } = split;
+    if (split.root === COMMENTS_ROOT) {
+        return format === 'rss'
+            ? { target: { kind: 'comments', document: undefined }, format, canonical }
+            : undefined;
+    }
+    const root = parseListingPath(split.root, bases, authors);
+    // Only a listing root has a feed, and only its first page: a feed is not
+    // paginated, so `/{base}/x/page/2/feed/` names nothing.
+    if (root !== undefined) {
+        return root.pageNumber === 0
+            ? { target: { kind: 'listing', subject: root }, format, canonical }
+            : undefined;
+    }
+    // Otherwise it may be a post's comments feed. Only a published post has one:
+    // a page never federates, so nothing in the fediverse can ever have replied
+    // to it, and a comments feed for it would be empty for ever.
+    if (format !== 'rss')
+        return undefined;
+    const document = publicDocumentAt(store, split.root);
+    if (document?.type !== 'post')
+        return undefined;
+    return { target: { kind: 'comments', document }, format, canonical };
+}
+/** Where the canonical URL of one feed is. */
+function feedTargetHref(target, format, bases) {
+    return target.kind === 'listing'
+        ? listingFeedHref(target.subject, format, bases)
+        : commentsFeedHref(target.document);
+}
+/**
+ * The format `?feed=` asked for, WordPress's pre-permalink spelling, or
+ * `undefined` when the request did not ask for a feed at all.
+ *
+ * `rss2` is what WordPress calls RSS 2.0 and `rss` its RSS 0.92, which this
+ * site does not serve and answers with the RSS 2.0 feed rather than a 404.
+ */
+function queryFeedFormat(c) {
+    const asked = c.req.query('feed');
+    if (asked === undefined)
+        return undefined;
+    if (asked === 'rss2' || asked === 'rss')
+        return 'rss';
+    return FEED_FORMATS.find((format) => FEED_SEGMENTS[format] === asked);
+}
+/**
+ * What the query string puts on a rendered post, if anything.
+ *
+ * A reader has no session, so the message after a submission travels in the
+ * URL and is read back here. An unknown value says nothing at all rather than
+ * being printed, so the query cannot be used to put words on somebody's post.
+ */
+function commentNotice(c, document) {
+    const notice = commentNoticeFor(c.req.query(COMMENT_NOTICE_PARAM));
+    return {
+        ...(notice === undefined ? {} : { commentNotice: notice }),
+        // And who a Reply link says the form is answering, which travels the same
+        // way and for the same reason: a reader has no session, and threading
+        // should not need a line of JavaScript.
+        ...commentReplyTarget({
+            admin: c.var.admin,
+            document,
+            id: c.req.query(COMMENT_REPLY_PARAM),
+        }),
+        // And the thank-you after a contact form was sent, which travels the same
+        // way for the same reason (TASK-56).
+        ...contactNotice(c),
+    };
+}
+/**
+ * The thank-you a contact form's redirect landed on, if the query names one.
+ *
+ * On the context rather than in the form's own state because the redirect is
+ * what makes a refresh harmless: the page is drawn fresh, and the query is the
+ * only thing that says a message went.
+ */
+function contactNotice(c) {
+    const notice = contactNoticeFor(c.req.query(CONTACT_NOTICE_PARAM));
+    return notice === undefined ? {} : { contactNotice: notice };
+}
 /** The representation an `Accept` header asked for, or `undefined` for a 406. */
 function selectFromAccept(c, available) {
     return selectRepresentation(c.req.header('accept'), available);
 }
-/** One document in one representation, or the 406 an impossible `Accept` earns. */
-function negotiateDocument(c, document, representation) {
+/**
+ * One document in one representation, or the 406 an impossible `Accept` earns.
+ *
+ * `href` is where it is being served, which is its own permalink everywhere
+ * but the front page: a page serving as the homepage is read at `/`, and the
+ * canonical URL, the alternates and the theme's own links have to say so.
+ */
+function negotiateDocument(c, document, representation, href = document.permalink) {
     if (representation === undefined) {
-        return notAcceptableResponse(encodePath(document.permalink), DOCUMENT_REPRESENTATIONS);
+        return notAcceptableResponse(encodePath(href), DOCUMENT_REPRESENTATIONS);
     }
+    // Who the request's session says is reading, when it says anybody
+    // (TASK-103). Only the HTML has a form on it, so only the HTML asks; the
+    // Markdown and JSON of a post are the same bytes for everybody.
+    const viewer = representation === 'html' ? signedInCommenter(c) : undefined;
     const body = representation === 'markdown'
         ? serializeDocument(document)
         : representation === 'json'
             ? documentJson(document, { baseUrl: c.var.config.baseUrl })
-            : c.var.renderer.renderDocument(document);
+            : // The thank-you after a comment was posted, which the redirect
+                // carried back as a query. It is the only thing about a document's
+                // HTML that the URL rather than the file decides.
+                // A document served at `/` is the site's front page, and the front
+                // page is the one place a theme may lay a page out differently.
+                href === '/'
+                    ? c.var.renderer.renderFrontPage(document, commentNotice(c, document), viewer)
+                    : c.var.renderer.renderDocument(document, commentNotice(c, document), viewer);
     // The theme can change without the document changing, and only the document
     // is hashed. While the watcher is on — a development server, where a template
     // edit lands mid-process — the HTML gets no validator rather than a stale one.
@@ -104,8 +396,17 @@ function negotiateDocument(c, document, representation) {
     return representationResponse({
         body,
         representation,
-        href: encodePath(document.permalink),
+        href: encodePath(href),
         available: DOCUMENT_REPRESENTATIONS,
+        // Where a webmention about this page is sent. It is a header rather than
+        // only a `<link>` because a sender is allowed to find the endpoint without
+        // parsing the page, and because the JSON and Markdown representations of a
+        // post have no head to put one in (TASK-51).
+        links: webmentionLinks(c),
+        // A page drawn for one named reader is that reader's page rather than the
+        // post, so nothing shared may hold it and it carries no validator. A page
+        // drawn for everybody is exactly what it has always been.
+        ...(viewer === undefined ? {} : { private: true }),
         ...(validated
             ? {
                 etag: representationEtag(representation, document.hash),
@@ -115,6 +416,19 @@ function negotiateDocument(c, document, representation) {
         conditional: conditionalHeaders(c),
     });
 }
+/**
+ * The `Link` header advertising this site's webmention endpoint, or none when
+ * the site does not take them.
+ *
+ * Read per request off the settings, so turning webmentions off on the
+ * settings screen takes the advertisement off the very next page rather than
+ * off the next restart.
+ */
+function webmentionLinks(c) {
+    if (!readSiteSettings(c.var.config.contentDir).webmentionsReceive)
+        return [];
+    return [`<${WEBMENTION_PATH}>; rel="webmention"`];
+}
 /** The conditional request headers, as the negotiator wants them. */
 function conditionalHeaders(c) {
     return {
@@ -131,55 +445,231 @@ function conditionalHeaders(c) {
  * itself rather than the path with a slash bolted on, so `/page/1` lands on
  * `/` in one hop instead of two.
  */
-function canonicalPath(c, pathname) {
+function canonicalPath(c, pathname, bases, authors) {
     if (pathname === '' || pathname.endsWith('/'))
         return undefined;
-    const target = canonicalTarget(c, `${pathname}/`);
+    const target = canonicalTarget(c, `${pathname}/`, bases, authors);
     if (target === undefined)
         return undefined;
     return `${target}${new URL(c.req.url).search}`;
 }
 /** Where a path with a trailing slash canonically lives, if anywhere. */
-function canonicalTarget(c, pathname) {
+function canonicalTarget(c, pathname, bases, authors) {
     const { store, renderer } = c.var;
+    // A feed first: `/feed`, `/feed/atom`, `/{base}/x/feed`, `/comments/feed`
+    // and a post's `{permalink}feed` all lead somewhere real, and `/feed/rss`
+    // leads to `/feed/` in the same one hop rather than to a second redirect.
+    const feedRequest = parseFeedPath(store, pathname, bases, authors);
+    if (feedRequest !== undefined) {
+        const { target } = feedRequest;
+        if (target.kind === 'listing' &&
+            target.subject.term !== undefined &&
+            countListing(store, target.subject) === 0) {
+            return undefined;
+        }
+        return feedTargetHref(target, feedRequest.format, bases);
+    }
+    const pages = frontPages(c);
     const document = store.getByPermalink(pathname);
     if (document !== undefined) {
-        return isPublicDocument(document) ? encodePath(pathname) : undefined;
+        if (!isPublicDocument(document))
+            return undefined;
+        // The homepage's own URL leads to `/` rather than to the redirect that
+        // leads to `/`: one hop, the way `/page/1` reaches `/` in one.
+        return document.path === pages.home?.path ? '/' : encodePath(pathname);
     }
-    const listing = parseListingPath(pathname);
+    const listing = listingRequestAt(pages, pathname, bases, authors);
     if (listing === undefined)
         return undefined;
-    // The home listing exists even with nothing on it; a tag archive does not.
-    const total = listing.tag === undefined ? store.counts().posts : store.countByTag(listing.tag);
-    if (listing.tag !== undefined && total === 0)
+    // The home listing and an author archive both exist with nothing on them; a
+    // taxonomy archive does not.
+    const total = countListing(store, listing);
+    if (listing.term !== undefined && total === 0)
         return undefined;
     const totalPages = Math.max(1, Math.ceil(total / renderer.pageSize()));
     if (listing.pageNumber >= totalPages)
         return undefined;
-    return listing.tag === undefined
-        ? homeHref(listing.pageNumber)
-        : tagHref(listing.tag, listing.pageNumber);
+    return listingHref(listing, listing.pageNumber, bases, listing.document?.permalink ?? '/');
+}
+/** {@link AuthorLookup} over one request's site. */
+function authorsOf(c) {
+    let users;
+    return (username) => {
+        users ??= listUsers(c.var.config.dataDir);
+        const user = users.find((candidate) => candidate.username === username);
+        return user === undefined ? undefined : { user, names: authorNames(users, user) };
+    };
+}
+/**
+ * WordPress's Reading choice, resolved for this request.
+ *
+ * The setting holds slugs, and a slug is only a front page while it names a
+ * page the public site would serve: one drafted, trashed or deleted since it
+ * was picked resolves to nothing, and a site whose homepage resolves to
+ * nothing is a site showing its latest posts again. That is the whole of the
+ * fallback — there is no state to repair and nothing to write — and it is why
+ * the pick is kept rather than cleared: publishing the page again puts the
+ * front page back.
+ */
+function frontPages(c) {
+    const { homepage, postsPage } = c.var.renderer.frontPageSlugs();
+    const home = publicPage(c, homepage);
+    return {
+        home,
+        // Without a homepage there is no posts page: the listing is already at
+        // `/`, and a second URL for it is exactly what this pair exists to avoid.
+        posts: home === undefined ? undefined : publicPage(c, postsPage),
+    };
 }
-/** `/`, `/page/N/`, `/tags/x/` and `/tags/x/page/N/`, as a listing to check. */
-function parseListingPath(pathname) {
+/**
+ * A path as the page of a listing it names, wherever the listing lives, or
+ * `undefined` when it names none.
+ *
+ * The posts page first, because its permalink is the listing's root on a site
+ * that has one; then the ordinary `/`, `/page/N/` and the taxonomy archives.
+ * A site whose homepage is a page and which named no posts page has no home
+ * listing at all, so nothing under the root is one either.
+ */
+function listingRequestAt(pages, pathname, bases, authors) {
+    const onPostsPage = pages.posts === undefined ? undefined : postsPageRequest(pages.posts, pathname);
+    if (onPostsPage !== undefined)
+        return onPostsPage;
+    const request = parseListingPath(pathname, bases, authors);
+    if (request === undefined)
+        return undefined;
+    // Only the home listing can be homeless. A tag archive and a person's
+    // archive have URLs of their own, whatever the Reading setting says `/` is.
+    if (request.term === undefined && request.author === undefined) {
+        if (listingRoot(pages) === undefined)
+            return undefined;
+    }
+    return request;
+}
+/**
+ * Where the post listing lives, or `undefined` when it has no page at all.
+ *
+ * A site showing its latest posts has it at `/`; a site with a posts page has
+ * it there; a site whose homepage is a page and which named no posts page has
+ * nowhere for it, which is WordPress's own answer and the reason its Reading
+ * screen offers the second pick. The feeds are unaffected either way.
+ */
+function listingRoot(pages) {
+    if (pages.posts !== undefined)
+        return pages.posts.permalink;
+    return pages.home === undefined ? '/' : undefined;
+}
+/** The published page one slug names, or `undefined`. */
+function publicPage(c, slug) {
+    if (slug === '')
+        return undefined;
+    const found = c.var.store.getBySlug(slug);
+    if (found?.type !== 'page')
+        return undefined;
+    return isPublicDocument(found, c.var.store.now()) ? found : undefined;
+}
+/**
+ * `{permalink}` or `{permalink}page/N/` as a page of the listing the posts
+ * page carries, or `undefined` when the path is neither.
+ *
+ * `page/1/` is not one: the first page of a listing lives at the listing's own
+ * URL, exactly as `/page/1/` collapses onto `/`, and the canonicalisation that
+ * says so is the ordinary one.
+ */
+function postsPageRequest(document, pathname) {
+    const { permalink } = document;
+    if (pathname === permalink)
+        return { term: undefined, pageNumber: 0, document };
+    if (!pathname.startsWith(permalink))
+        return undefined;
+    const segments = pathname
+        .slice(permalink.length)
+        .split('/')
+        .filter((segment) => segment !== '');
+    if (segments.length !== 2 || segments[0] !== PAGE_SEGMENT || !pathname.endsWith('/')) {
+        return undefined;
+    }
+    const page = pageIndex(segments[1]);
+    return page === undefined ? undefined : { term: undefined, pageNumber: page, document };
+}
+/**
+ * `/`, `/page/N/`, `/author/{username}/` and `/author/{username}/page/N/`, and
+ * either taxonomy's `/{base}/x/` and `/{base}/x/page/N/`, as a listing to
+ * check.
+ */
+function parseListingPath(pathname, bases, authors) {
     const segments = pathname.split('/').filter((segment) => segment !== '');
     if (segments.length === 0)
-        return { tag: undefined, pageNumber: 0 };
+        return { term: undefined, pageNumber: 0 };
     if (segments[0] === PAGE_SEGMENT && segments.length === 2) {
         const page = pageIndex(segments[1]);
-        return page === undefined ? undefined : { tag: undefined, pageNumber: page };
+        return page === undefined ? undefined : { term: undefined, pageNumber: page };
+    }
+    // One person's archive, at the path decision-14 reserves. A username nobody
+    // has is not a listing at all, which is what turns it into one 404 rather
+    // than an empty archive.
+    const byAuthor = parseAuthorPath(pathname);
+    if (byAuthor !== undefined) {
+        const found = authors(byAuthor.username);
+        return found === undefined ? undefined : { author: found, pageNumber: byAuthor.pageNumber };
     }
-    if (segments[0] !== TAG_SEGMENT || segments[1] === undefined)
+    const taxonomy = taxonomyForSegment(segments[0], bases);
+    if (taxonomy === undefined || segments[1] === undefined)
         return undefined;
-    const tag = segments[1];
+    const term = { taxonomy, term: segments[1] };
     if (segments.length === 2)
-        return { tag, pageNumber: 0 };
+        return { term, pageNumber: 0 };
     if (segments.length === 4 && segments[2] === PAGE_SEGMENT) {
         const page = pageIndex(segments[3]);
-        return page === undefined ? undefined : { tag, pageNumber: page };
+        return page === undefined ? undefined : { term, pageNumber: page };
     }
     return undefined;
 }
+/** How many published documents a listing holds. */
+function countListing(store, subject) {
+    const { term, author } = subject;
+    if (author !== undefined)
+        return store.countByAuthor(author.names);
+    if (term === undefined)
+        return store.counts().posts;
+    return term.taxonomy === 'tag' ? store.countByTag(term.term) : store.countByCategory(term.term);
+}
+/** One page of a listing's documents, newest first. */
+function listListing(store, subject, paging) {
+    const { term, author } = subject;
+    if (author !== undefined)
+        return store.listByAuthor(author.names, paging);
+    if (term === undefined)
+        return store.listPosts(paging);
+    return term.taxonomy === 'tag'
+        ? store.listByTag(term.term, paging)
+        : store.listByCategory(term.term, paging);
+}
+/**
+ * The URL of a page of a listing: the home archive's, or a taxonomy's.
+ *
+ * `root` is where the home listing lives — `/`, or the posts page's permalink
+ * on a site that has one — so every link the listing draws, every canonical
+ * redirect and the sitemap all spell its pages the same way.
+ */
+function listingHref(subject, index, bases, root = '/') {
+    const { term, author } = subject;
+    if (author !== undefined)
+        return authorHref(author.user.username, index);
+    return term === undefined ? listingPageHref(root, index) : termHref(term, index, bases);
+}
+/**
+ * The URL of one format of one listing's feed, whichever kind of listing it is.
+ *
+ * The site's and a taxonomy's hang off the listing root; a person's hangs off
+ * their archive, which decision-14 keeps at WordPress's `/author/{username}/`
+ * rather than under the listing machinery, because that root is also an actor
+ * id.
+ */
+function listingFeedHref(subject, format, bases) {
+    return subject.author === undefined
+        ? feedHref(subject.term, format, bases)
+        : authorFeedHref(subject.author.user.username, format);
+}
 /**
  * A `/page/N/` segment as a zero-based index. `page/1` parses to the first
  * page, which lives at the listing root, so canonicalisation collapses it.
@@ -193,35 +683,56 @@ function pageIndex(segment) {
 /** One page of a listing, in whichever representation the request settled on. */
 function listing(c, request, representation = selectFromAccept(c, LISTING_REPRESENTATIONS)) {
     const { store, renderer } = c.var;
-    const { tag } = request;
+    const { term, author } = request;
     const size = renderer.pageSize();
-    // A tag archive only exists while something carries the tag; the home
-    // listing exists even with nothing on it.
-    const total = tag === undefined ? store.counts().posts : store.countByTag(tag);
-    if (tag !== undefined && total === 0)
+    const bases = renderer.taxonomyBases();
+    // Where this listing's pages live: the posts page's permalink when the site
+    // has one, and the site root otherwise.
+    const root = request.document?.permalink ?? '/';
+    // WordPress served every feed as a query on the listing before it served one
+    // at a path, and the links are still out there. `/?feed=rss2` and
+    // `/{base}/x/?feed=atom` land on the feed the listing now has.
+    const asked = queryFeedFormat(c);
+    if (asked !== undefined)
+        return c.redirect(listingFeedHref(request, asked, bases), 301);
+    // A taxonomy archive only exists while something carries the term; the home
+    // listing and a person's archive exist even with nothing on them.
+    const total = countListing(store, request);
+    if (term !== undefined && total === 0)
         return notFound(c);
-    const href = tag === undefined ? homeHref(request.pageNumber) : tagHref(tag, request.pageNumber);
+    const href = listingHref(request, request.pageNumber, bases, root);
     const pagination = paginate({
         total,
         size,
         pageNumber: request.pageNumber,
-        hrefForPage: (index) => (tag === undefined ? homeHref(index) : tagHref(tag, index)),
+        hrefForPage: (index) => listingHref(request, index, bases, root),
     });
     if (request.pageNumber >= pagination.totalPages)
         return notFound(c);
     if (representation === undefined)
         return notAcceptableResponse(href, LISTING_REPRESENTATIONS);
     const paging = { limit: size, offset: offsetForPage(request.pageNumber, size) };
-    const documents = tag === undefined ? store.listPosts(paging) : store.listByTag(tag, paging);
+    const documents = listListing(store, request, paging);
     const full = wantsFullDocuments(c);
     const body = representation === 'json'
         ? documents.map((document) => documentJson(document, { baseUrl: c.var.config.baseUrl, body: full }))
         : renderer.renderListing({
-            title: tag ?? renderer.site().title,
+            // An author archive is headed by the person, a taxonomy archive by
+            // the term, the posts page by its own title the way any page is, and
+            // the home listing by the site's.
+            title: term?.term ??
+                author?.user.profile?.displayName ??
+                author?.user.username ??
+                request.document?.title ??
+                renderer.site().title,
             url: href,
             documents,
             pagination,
-            ...(tag === undefined ? {} : { tag, template: TEMPLATES.tag }),
+            // And its words go above the posts, which is what a posts page is
+            // for: a theme prints `{{ content | safe }}` over the list.
+            ...(request.document === undefined ? {} : { document: request.document }),
+            ...(term === undefined ? {} : taxonomyContext(term)),
+            ...(author === undefined ? {} : authorArchiveContext(author)),
         });
     const validated = representation !== 'html' || !c.var.config.watch;
     return representationResponse({
@@ -238,6 +749,59 @@ function listing(c, request, representation = selectFromAccept(c, LISTING_REPRES
         conditional: conditionalHeaders(c),
     });
 }
+/**
+ * One page of search results, in whichever representation the request settled
+ * on.
+ *
+ * What it finds is what the listings would show: the index holds the search to
+ * published, untrashed, already-due documents, so a draft cannot be found by
+ * its words any more than it can be found by its URL. A search with no query
+ * is the page with the form on it and nothing found, rather than an error,
+ * because that is where a link to the search from a theme lands. A page past
+ * the last one 404s, as it does on a listing.
+ */
+function search(c, representation) {
+    const { store, renderer, config } = c.var;
+    const query = searchQuery(c.req.query(SEARCH_QUERY_PARAM));
+    const pageNumber = searchPageIndex(c.req.query(SEARCH_PAGE_PARAM));
+    if (pageNumber === undefined)
+        return notFound(c);
+    const size = renderer.pageSize();
+    const pagination = paginate({
+        total: store.countSearch(query),
+        size,
+        pageNumber,
+        hrefForPage: (index) => searchHref(query, index),
+    });
+    if (pageNumber >= pagination.totalPages)
+        return notFound(c);
+    const href = searchHref(query, pageNumber);
+    if (representation === undefined)
+        return notAcceptableResponse(href, LISTING_REPRESENTATIONS);
+    const hits = store.search(query, { limit: size, offset: offsetForPage(pageNumber, size) });
+    const documents = hits.map((hit) => hit.document);
+    const full = wantsFullDocuments(c);
+    const body = representation === 'json'
+        ? searchJson({ query, hits, pagination, baseUrl: config.baseUrl, body: full })
+        : renderer.renderSearch({ query, url: href, hits, pagination });
+    const validated = representation !== 'html' || !config.watch;
+    return representationResponse({
+        body,
+        representation,
+        href,
+        // The alternates are spelled here rather than by the negotiator, because
+        // they carry the query and a listing's never have one to carry.
+        available: [representation],
+        links: LISTING_REPRESENTATIONS.filter((other) => other !== representation).map((other) => `<${searchHref(query, pageNumber, other)}>; rel="alternate"; type="${MEDIA_TYPES[other]}"`),
+        ...(validated
+            ? {
+                etag: representationEtag(representation, listingFingerprint(href, pagination, documents, full)),
+                lastModified: latestModified(documents),
+            }
+            : {}),
+        conditional: conditionalHeaders(c),
+    });
+}
 /**
  * Whether the request asked for whole documents rather than summaries.
  *
@@ -264,31 +828,155 @@ function listingFingerprint(href, pagination, documents, full) {
     ].join('\n');
 }
 /**
- * The site's feed, or one tag's, in one format.
+ * The site's feed, or one archive's, in one format.
  *
- * A tag with nothing published under it 404s exactly as its archive does: a
+ * A term with nothing published under it 404s exactly as its archive does: a
  * feed reader should be told the URL is wrong rather than handed an empty feed
  * it will poll forever.
  */
-function feed(c, format, tag) {
+function feed(c, format, subject) {
     const { store, renderer, config } = c.var;
     const site = renderer.site();
-    if (tag !== undefined && store.countByTag(tag) === 0)
-        return notFound(c);
-    const documents = tag === undefined
-        ? store.listPosts({ limit: feedSize(site) })
-        : store.listByTag(tag, { limit: feedSize(site) });
-    const href = tag === undefined ? homeHref(0) : tagHref(tag, 0);
+    const bases = renderer.taxonomyBases();
+    const { term, author } = subject;
+    if (term !== undefined && countListing(store, subject) === 0) {
+        // A subscriber to a renamed archive's feed follows it to the new one
+        // rather than being dropped, exactly as a reader of the archive does.
+        const moved = movedTerm(c, term);
+        if (moved === undefined)
+            return notFound(c);
+        return c.redirect(feedHref(moved, format, bases), 301);
+    }
+    const documents = listListing(store, subject, { limit: feedSize(site) });
+    const href = listingHref(subject, 0, bases);
+    // What the feed calls itself: the site, and then the term or the person it
+    // is about, so a reader subscribed to several of a site's feeds can tell
+    // them apart in a list.
+    const about = term?.term ?? author?.user.profile?.displayName ?? author?.user.username;
     const source = {
         site,
         documents,
-        title: tag === undefined ? site.title : `${site.title}: ${tag}`,
+        title: about === undefined ? site.title : `${site.title}: ${about}`,
         href,
-        feedHref: feedHref(tag, format),
+        feedHref: listingFeedHref(subject, format, bases),
         baseUrl: config.baseUrl,
+        // Only RSS carries the comment pointers; the other two formats have no
+        // vocabulary for them, and counting for a feed that cannot say the number
+        // would be a query per item for nothing.
+        ...(format === 'rss' ? { commentCounts: c.var.conversation.counts(documents) } : {}),
     };
     return feedResponse({ format, source, conditional: conditionalHeaders(c) });
 }
+/**
+ * A post's comments, or the whole site's, as RSS 2.0.
+ *
+ * A post with no replies answers an empty feed rather than a 404: it exists,
+ * and a reader that subscribed before anybody answered should keep polling.
+ * Only a permalink that is no published post 404s, which is the ordinary
+ * document lookup rather than anything this feed decides.
+ */
+function comments(c, document) {
+    const { conversation, renderer, config } = c.var;
+    const site = renderer.site();
+    const limit = feedSize(site);
+    // The same reading the page is drawn from, which is the point: a subscriber
+    // to a post's comments and a reader who scrolls to the bottom of it are
+    // looking at one conversation. A post's own feed hands the builder entries
+    // with no post attached, and so its items name no post: there, every item
+    // answers the same one.
+    const found = feedComments(document === undefined ? conversation.latest(limit) : spokenIn(conversation.thread(document)), { baseUrl: config.baseUrl, limit });
+    const source = {
+        site,
+        comments: found,
+        title: document === undefined
+            ? `${site.title}: comments`
+            : `${COMMENTS_TITLE_PREFIX}${document.title}`,
+        href: document?.permalink ?? '/',
+        feedHref: commentsFeedHref(document),
+        baseUrl: config.baseUrl,
+    };
+    return commentsFeedResponse(source, conditionalHeaders(c));
+}
+/**
+ * The sitemap, or one of its children once the site is too big for one file.
+ *
+ * A child that does not exist 404s through the theme like any other missing
+ * URL: `/sitemap-1.xml` names nothing at all on a site whose whole sitemap
+ * fits, and a crawler asking for one should be told so rather than handed an
+ * empty file it would come back to.
+ */
+function sitemap(c, page) {
+    const response = sitemapResponse({
+        urls: sitemapUrls(c),
+        baseUrl: c.var.config.baseUrl,
+        page,
+        conditional: conditionalHeaders(c),
+    });
+    return response ?? notFound(c);
+}
+/**
+ * Every URL the public site publishes, in the order the sitemap lists them:
+ * the home archive's pages, then every post and page, then every tag archive
+ * and every category archive with their own pages.
+ *
+ * Everything here is spelled by the same functions the site's own links go
+ * through — {@link homeHref}, the permalinks, {@link termHref} — and drawn
+ * from the same queries the listings use, so a sitemap can never advertise a
+ * URL the site does not serve. Nothing hidden reaches it either: the index's
+ * public queries already exclude drafts, the trash and posts whose date has
+ * not arrived, and {@link isPublicDocument} is asked again with the store's
+ * own clock so the two answers cannot drift apart.
+ */
+function sitemapUrls(c) {
+    const { store, renderer } = c.var;
+    const now = store.now();
+    const size = renderer.pageSize();
+    const bases = renderer.taxonomyBases();
+    const pages = frontPages(c);
+    const urls = [];
+    /** The pages of one listing, each dated by the newest document on it. */
+    const listingPages = (documents, hrefForPage) => {
+        const totalPages = Math.max(1, Math.ceil(documents.length / size));
+        for (let index = 0; index < totalPages; index += 1) {
+            const onPage = documents.slice(index * size, (index + 1) * size);
+            urls.push({ loc: hrefForPage(index), lastmod: latestModified(onPage) });
+        }
+    };
+    const posts = store.listPosts().filter((document) => isPublicDocument(document, now));
+    // The listing's pages, wherever it lives: under `/`, under the posts page,
+    // or nowhere at all on a site whose homepage is a page and which named no
+    // posts page. Its first page is that root, which is why `/` and the posts
+    // page's own URL are listed here rather than with the documents.
+    const root = listingRoot(pages);
+    if (root !== undefined)
+        listingPages(posts, (index) => listingPageHref(root, index));
+    // And `/` itself when it is a page rather than the listing: the front page
+    // is published at one URL, and it is this one.
+    if (pages.home !== undefined)
+        urls.push({ loc: '/', lastmod: lastModifiedOf(pages.home) });
+    const documents = store
+        .listAll({ type: 'page', draft: false, trashed: false, scheduled: false })
+        .filter((document) => isPublicDocument(document, now));
+    for (const document of [...posts, ...documents]) {
+        // The homepage answers at `/` and redirects from its own permalink, and
+        // the posts page is the listing's first page, which is already listed:
+        // neither is advertised twice.
+        if (document.path === pages.home?.path || document.path === pages.posts?.path)
+            continue;
+        urls.push({ loc: document.permalink, lastmod: lastModifiedOf(document) });
+    }
+    // A taxonomy archive exists only while something public carries the term,
+    // which is exactly what these two queries report.
+    for (const { tag } of store.listTags()) {
+        const term = { taxonomy: 'tag', term: tag };
+        listingPages(store.listByTag(tag), (index) => termHref(term, index, bases));
+    }
+    for (const { category } of store.listCategories()) {
+        const term = { taxonomy: 'category', term: category };
+        listingPages(store.listByCategory(category), (index) => termHref(term, index, bases));
+    }
+    return urls;
+}
 /**
  * A file from the theme's `static/` directory, the site's copy first.
  *
@@ -297,7 +985,7 @@ function feed(c, format, tag) {
  */
 function themeAsset(c) {
     const relative = requestPath(c).slice(THEME_ASSET_PREFIX.length);
-    const asset = findThemeAsset(relative, themeSearchPath(c.var.config.themeDir));
+    const asset = findThemeAsset(relative, c.var.renderer.themeDirs());
     if (asset === undefined)
         return notFound(c);
     if (matchesEtag(c.req.header('if-none-match'), asset.etag)) {
@@ -305,6 +993,38 @@ function themeAsset(c) {
     }
     return themeAssetResponse(asset);
 }
+/**
+ * One derived copy of an upload, generated on the spot if it is not there.
+ *
+ * The directory these live in is disposable (decision-9): deleting it costs
+ * the next reader of each picture one encode and costs the site nothing else,
+ * which is what this handler is for. A width or a format the site does not
+ * offer is a 404 and encodes nothing, so the URL space cannot be used to make
+ * the server work.
+ *
+ * The cache lifetime is the uploads' own, and it is honest for the same
+ * reason: a derived URL names one width of one file, and the file it was
+ * derived from is never overwritten.
+ */
+async function imageVariant(c) {
+    const relative = requestPath(c).slice(VARIANT_ASSET_PREFIX.length);
+    const asset = await findImageVariant(c.var.config, decodeVariantPath(relative));
+    if (asset === undefined)
+        return notFound(c);
+    const options = { maxAge: UPLOAD_ASSET_MAX_AGE };
+    return matchesEtag(c.req.header('if-none-match'), asset.etag)
+        ? assetNotModified(asset, options)
+        : assetResponse(asset, options);
+}
+/** A request path as a path on disk: percent-encoding off, segment by segment. */
+function decodeVariantPath(relative) {
+    try {
+        return relative.split('/').map(decodeURIComponent).join('/');
+    }
+    catch {
+        return relative;
+    }
+}
 /**
  * A file from `content/uploads/`, at the URL Eleventy's passthrough copy puts
  * it at.
@@ -327,16 +1047,57 @@ function upload(c) {
 }
 /** The URL of a page of the home listing, by zero-based index. */
 export function homeHref(index) {
-    return index === 0 ? '/' : `/${PAGE_SEGMENT}/${String(index + 1)}/`;
+    return listingPageHref('/', index);
 }
-/** The URL of a page of a tag archive, by zero-based index. */
-export function tagHref(tag, index) {
-    const root = `/${TAG_SEGMENT}/${encodeURIComponent(tag)}/`;
+/**
+ * The URL of a page of the listing that hangs off `root`, by zero-based index.
+ *
+ * The first page is the root itself: `/` or the posts page's own permalink,
+ * which is what makes a posts page one URL rather than two.
+ */
+export function listingPageHref(root, index) {
     return index === 0 ? root : `${root}${PAGE_SEGMENT}/${String(index + 1)}/`;
 }
-/** The URL of a feed: the whole archive's, or one tag's. */
-export function feedHref(tag, format) {
-    return tag === undefined ? `/${FEED_FILES[format]}` : `${tagHref(tag, 0)}${FEED_FILES[format]}`;
+/**
+ * What the theme is told about a taxonomy archive: the term under the name of
+ * its taxonomy, and the layout that taxonomy uses.
+ */
+function taxonomyContext(term) {
+    return term.taxonomy === 'tag'
+        ? { tag: term.term, template: TEMPLATES.tag }
+        : { category: term.term, template: TEMPLATES.category };
+}
+/**
+ * What the theme is told about an author archive: the person's profile under
+ * `author`, the same shape a post's byline gets, and the layout an archive of
+ * theirs uses.
+ *
+ * One shape for both places is the point. A theme that knows how to print a
+ * byline already knows how to print the heading of the archive that byline
+ * links to, and neither has to ask whether the site has users.
+ */
+function authorArchiveContext(author) {
+    return { author: profileContext(author.user), template: TEMPLATES.author };
+}
+/**
+ * The URL of a feed: the whole site's, or one taxonomy archive's.
+ *
+ * WordPress's layout, so a subscriber of a migrated site keeps polling the URL
+ * they already hold: `/feed/` is RSS 2.0, `/feed/atom/` and `/feed/json/` are
+ * its siblings, and an archive's feeds hang off the archive.
+ */
+export function feedHref(term, format, bases) {
+    return feedPathUnder(listingHref({ term }, 0, bases), format);
+}
+/**
+ * Where the site's comments feed lives, and where one post's does.
+ *
+ * WordPress's URLs, so a site migrated from it keeps both: the whole site's
+ * comments at `/comments/feed/`, and a post's under its own permalink. There
+ * is one format, RSS 2.0, because that is what a comments feed is read in.
+ */
+export function commentsFeedHref(document) {
+    return commentsFeedPath(document === undefined ? undefined : encodePath(document.permalink));
 }
 /** The theme's 404 page. */
 export function notFound(c) {