X-Git-Url: http://www.git.stargrave.org/?a=blobdiff_plain;f=lib%2FPublicInbox%2FWWW.pm;h=85cb234b4d56351589bbbf8d633893e7498a17b6;hb=b836c9d3a2053fa97a98cc38e1657d3266173c7e;hp=1d1a0c9d468fae37e509e6a0a4c9ddcb1cc232c4;hpb=efdeec171349d5793f5af78c31f4b2d0c80a7ec8;p=public-inbox.git

diff --git a/lib/PublicInbox/WWW.pm b/lib/PublicInbox/WWW.pm
index 1d1a0c9d..534ee028 100644
--- a/lib/PublicInbox/WWW.pm
+++ b/lib/PublicInbox/WWW.pm
@@ -1,278 +1,637 @@
-# Copyright (C) 2014, Eric Wong <normalperson@yhbt.net> and all contributors
-# License: AGPLv3 or later (https://www.gnu.org/licenses/agpl-3.0.txt)
+# Copyright (C) 2014-2020 all contributors <meta@public-inbox.org>
+# License: AGPL-3.0+ <https://www.gnu.org/licenses/agpl-3.0.txt>
+#
+# Main web interface for mailing list archives
 #
 # We focus on the lowest common denominators here:
-# - targeted at text-only console browsers (lynx, w3m, etc..)
+# - targeted at text-only console browsers (w3m, links, etc..)
 # - Only basic HTML, CSS only for line-wrapping <pre> text content for GUIs
+#   and diff/syntax-highlighting (optional)
 # - No JavaScript, graphics or icons allowed.
 # - Must not rely on static content
 # - UTF-8 is only for user-content, 7-bit US-ASCII for us
 package PublicInbox::WWW;
-use 5.008;
+use 5.010_001;
 use strict;
 use warnings;
+use bytes (); # only for bytes::length
 use PublicInbox::Config;
-use URI::Escape qw(uri_escape_utf8 uri_unescape);
-use constant SSOMA_URL => 'http://ssoma.public-inbox.org/';
-use constant PI_URL => 'http://public-inbox.org/';
-our $LISTNAME_RE = qr!\A/([\w\.\-]+)!;
-our $pi_config;
-BEGIN {
-	$pi_config = PublicInbox::Config->new;
+use PublicInbox::Hval;
+use URI::Escape qw(uri_unescape);
+use PublicInbox::MID qw(mid_escape);
+use PublicInbox::GitHTTPBackend;
+use PublicInbox::UserContent;
+use PublicInbox::WwwStatic qw(r path_info_raw);
+
+# TODO: consider a routing tree now that we have more endpoints:
+our $INBOX_RE = qr!\A/([\w\-][\w\.\-]*)!;
+our $MID_RE = qr!([^/]+)!;
+our $END_RE = qr!(T/|t/|t\.mbox(?:\.gz)?|t\.atom|raw|)!;
+our $ATTACH_RE = qr!([0-9][0-9\.]*)-($PublicInbox::Hval::FN)!;
+our $OID_RE = qr![a-f0-9]{7,40}!;
+
+sub new {
+	my ($class, $pi_config) = @_;
+	$pi_config ||= PublicInbox::Config->new;
+	bless { pi_config => $pi_config }, $class;
 }
 
+# backwards compatibility, do not use
 sub run {
-	my ($cgi, $method) = @_;
-	my %ctx;
-	if ($method !~ /\AGET|HEAD\z/) {
-		return r(405, 'Method Not Allowed');
+	my ($req, $method) = @_;
+	PublicInbox::WWW->new->call($req->env);
+}
+
+sub call {
+	my ($self, $env) = @_;
+	my $ctx = { env => $env, www => $self };
+
+	# we don't care about multi-value
+	%{$ctx->{qp}} = map {
+		utf8::decode($_);
+		tr/+/ /;
+		my ($k, $v) = split('=', $_, 2);
+		$v = uri_unescape($v // '');
+		# none of the keys we care about will need escaping
+		$k => $v;
+	} split(/[&;]+/, $env->{QUERY_STRING});
+
+	my $path_info = path_info_raw($env);
+	my $method = $env->{REQUEST_METHOD};
+
+	if ($method eq 'POST') {
+		if ($path_info =~ m!$INBOX_RE/(?:(?:git/)?([0-9]+)(?:\.git)?/)?
+					(git-upload-pack)\z!x) {
+			my ($epoch, $path) = ($2, $3);
+			return invalid_inbox($ctx, $1) ||
+				serve_git($ctx, $epoch, $path);
+		} elsif ($path_info =~ m!$INBOX_RE/!o) {
+			return invalid_inbox($ctx, $1) || mbox_results($ctx);
+		}
+	}
+	elsif ($method !~ /\A(?:GET|HEAD)\z/) {
+		return r(405);
 	}
-	my $path_info = $cgi->path_info;
 
 	# top-level indices and feeds
-	if ($path_info eq '/') {
-		r404();
-	} elsif ($path_info =~ m!$LISTNAME_RE\z!o) {
-		invalid_list(\%ctx, $1) || redirect_list_index(\%ctx, $cgi);
-	} elsif ($path_info =~ m!$LISTNAME_RE(?:/|/index\.html)?\z!o) {
-		invalid_list(\%ctx, $1) || get_index(\%ctx, $cgi, 0);
-	} elsif ($path_info =~ m!$LISTNAME_RE/atom\.xml\z!o) {
-		invalid_list(\%ctx, $1) || get_atom(\%ctx, $cgi, 0);
-
-	# single-message pages
-	} elsif ($path_info =~ m!$LISTNAME_RE/m/(\S+)\.txt\z!o) {
-		invalid_list_mid(\%ctx, $1, $2) || get_mid_txt(\%ctx, $cgi);
-	} elsif ($path_info =~ m!$LISTNAME_RE/m/(\S+)\.html\z!o) {
-		invalid_list_mid(\%ctx, $1, $2) || get_mid_html(\%ctx, $cgi);
-
-	# full-message page
-	} elsif ($path_info =~ m!$LISTNAME_RE/f/(\S+)\.html\z!o) {
-		invalid_list_mid(\%ctx, $1, $2) || get_full_html(\%ctx, $cgi);
-
-	# convenience redirects, order matters
-	} elsif ($path_info =~ m!$LISTNAME_RE/(?:m|f)/(\S+)\z!o) {
-		invalid_list_mid(\%ctx, $1, $2) || redirect_mid(\%ctx, $cgi);
+	if ($path_info eq '/' || $path_info eq '/manifest.js.gz') {
+		www_listing($self)->call($env);
+	} elsif ($path_info =~ m!$INBOX_RE\z!o) {
+		invalid_inbox($ctx, $1) || r301($ctx, $1);
+	} elsif ($path_info =~ m!$INBOX_RE(?:/|/index\.html)?\z!o) {
+		invalid_inbox($ctx, $1) || get_index($ctx);
+	} elsif ($path_info =~ m!$INBOX_RE/(?:atom\.xml|new\.atom)\z!o) {
+		invalid_inbox($ctx, $1) || get_atom($ctx);
+	} elsif ($path_info =~ m!$INBOX_RE/new\.html\z!o) {
+		invalid_inbox($ctx, $1) || get_new($ctx);
+	} elsif ($path_info =~ m!$INBOX_RE/description\z!o) {
+		get_description($ctx, $1);
+	} elsif ($path_info =~ m!$INBOX_RE/(?:(?:git/)?([0-9]+)(?:\.git)?/)?
+				($PublicInbox::GitHTTPBackend::ANY)\z!ox) {
+		my ($epoch, $path) = ($2, $3);
+		invalid_inbox($ctx, $1) || serve_git($ctx, $epoch, $path);
+	} elsif ($path_info =~ m!$INBOX_RE/([a-zA-Z0-9_\-]+).mbox\.gz\z!o) {
+		serve_mbox_range($ctx, $1, $2);
+	} elsif ($path_info =~ m!$INBOX_RE/$MID_RE/$END_RE\z!o) {
+		msg_page($ctx, $1, $2, $3);
+
+	} elsif ($path_info =~ m!$INBOX_RE/$MID_RE/$ATTACH_RE\z!o) {
+		my ($idx, $fn) = ($3, $4);
+		invalid_inbox_mid($ctx, $1, $2) || get_attach($ctx, $idx, $fn);
+	# in case people leave off the trailing slash:
+	} elsif ($path_info =~ m!$INBOX_RE/$MID_RE/(T|t)\z!o) {
+		my ($inbox, $mid_ue, $suffix) = ($1, $2, $3);
+		$suffix .= $suffix =~ /\A[tT]\z/ ? '/#u' : '/';
+		r301($ctx, $inbox, $mid_ue, $suffix);
+
+	} elsif ($path_info =~ m!$INBOX_RE/$MID_RE/R/?\z!o) {
+		my ($inbox, $mid_ue) = ($1, $2);
+		r301($ctx, $inbox, $mid_ue, '#R');
+
+	} elsif ($path_info =~ m!$INBOX_RE/$MID_RE/f/?\z!o) {
+		r301($ctx, $1, $2);
+	} elsif ($path_info =~ m!$INBOX_RE/_/text(?:/(.*))?\z!o) {
+		get_text($ctx, $1, $2);
+	} elsif ($path_info =~ m!$INBOX_RE/([a-zA-Z0-9_\-\.]+)\.css\z!o) {
+		get_css($ctx, $1, $2);
+	} elsif ($path_info =~ m!$INBOX_RE/manifest\.js\.gz\z!o) {
+		get_inbox_manifest($ctx, $1, $2);
+	} elsif ($path_info =~ m!$INBOX_RE/($OID_RE)/s/\z!o) {
+		get_vcs_object($ctx, $1, $2);
+	} elsif ($path_info =~ m!$INBOX_RE/($OID_RE)/s/
+				($PublicInbox::Hval::FN)\z!ox) {
+		get_vcs_object($ctx, $1, $2, $3);
+	} elsif ($path_info =~ m!$INBOX_RE/($OID_RE)/s\z!o) {
+		r301($ctx, $1, $2, 's/');
+	# convenience redirects order matters
+	} elsif ($path_info =~ m!$INBOX_RE/([^/]{2,})\z!o) {
+		r301($ctx, $1, $2);
 
 	} else {
-		r404();
+		legacy_redirects($ctx, $path_info);
 	}
 }
 
 # for CoW-friendliness, MOOOOO!
 sub preload {
+	my ($self) = @_;
+	require PublicInbox::ExtMsg;
 	require PublicInbox::Feed;
 	require PublicInbox::View;
-	require PublicInbox::Thread;
-	require Email::MIME;
-	require Digest::SHA;
-	require POSIX;
-	require XML::Atom::SimpleFeed;
+	require PublicInbox::SearchThread;
+	require PublicInbox::MIME;
+	require PublicInbox::Mbox;
+	require PublicInbox::ViewVCS;
+	require PublicInbox::WwwText;
+	require PublicInbox::WwwAttach;
+	eval {
+		require PublicInbox::Search;
+		PublicInbox::Search::load_xapian();
+	};
+	foreach (qw(PublicInbox::SearchView
+			PublicInbox::MboxGz
+			PublicInbox::NewsWWW)) {
+		eval "require $_;";
+	}
+	if (ref($self)) {
+		$self->cgit;
+		$self->stylesheets_prepare($_) for ('', '../', '../../');
+		$self->www_listing;
+	}
 }
 
 # private functions below
 
-sub r404 { r(404, 'Not Found') }
+sub r404 {
+	my ($ctx) = @_;
+	if ($ctx && $ctx->{mid}) {
+		require PublicInbox::ExtMsg;
+		return PublicInbox::ExtMsg::ext_msg($ctx);
+	}
+	r(404);
+}
 
-# simple response for errors
-sub r { [ $_[0], ['Content-Type' => 'text/plain'], [ join(' ', @_, "\n") ] ] }
+sub news_cgit_fallback ($) {
+	my ($ctx) = @_;
+	my $www = $ctx->{www};
+	my $env = $ctx->{env};
+	my $res = $www->news_www->call($env);
+	$res->[0] == 404 ? $www->cgit->call($env) : $res;
+}
 
 # returns undef if valid, array ref response if invalid
-sub invalid_list {
-	my ($ctx, $listname) = @_;
-	my $git_dir = $pi_config->get($listname, "mainrepo");
-	if (defined $git_dir) {
-		$ctx->{git_dir} = $git_dir;
-		$ctx->{listname} = $listname;
+sub invalid_inbox ($$) {
+	my ($ctx, $inbox) = @_;
+	my $ibx = $ctx->{www}->{pi_config}->lookup_name($inbox);
+	if (defined $ibx) {
+		$ctx->{-inbox} = $ibx;
 		return;
 	}
-	r404();
+
+	# sometimes linkifiers (not ours!) screw up automatic link
+	# generation and link things intended for nntp:// to https?://,
+	# so try to infer links and redirect them to the appropriate
+	# list URL.
+	news_cgit_fallback($ctx);
 }
 
 # returns undef if valid, array ref response if invalid
-sub invalid_list_mid {
-	my ($ctx, $listname, $mid) = @_;
-	my $ret = invalid_list($ctx, $listname, $mid);
-	$ctx->{mid} = uri_unescape($mid) unless $ret;
-	$ret;
+sub invalid_inbox_mid {
+	my ($ctx, $inbox, $mid_ue) = @_;
+	my $ret = invalid_inbox($ctx, $inbox);
+	return $ret if $ret;
+
+	my $mid = $ctx->{mid} = uri_unescape($mid_ue);
+	my $ibx = $ctx->{-inbox};
+	if ($mid =~ m!\A([a-f0-9]{2})([a-f0-9]{38})\z!) {
+		my ($x2, $x38) = ($1, $2);
+		# this is horrifically wasteful for legacy URLs:
+		my $str = $ctx->{-inbox}->msg_by_path("$x2/$x38") or return;
+		require Email::Simple;
+		my $s = Email::Simple->new($str);
+		$mid = PublicInbox::MID::mid_clean($s->header('Message-ID'));
+		return r301($ctx, $inbox, mid_escape($mid));
+	}
+	undef;
 }
 
-# /$LISTNAME/atom.xml                       -> Atom feed, includes replies
+# /$INBOX/new.atom                     -> Atom feed, includes replies
 sub get_atom {
-	my ($ctx, $cgi, $top) = @_;
+	my ($ctx) = @_;
 	require PublicInbox::Feed;
-	[ 200, [ 'Content-Type' => 'application/xml' ],
-	  [ PublicInbox::Feed->generate({
-			git_dir => $ctx->{git_dir},
-			listname => $ctx->{listname},
-			pi_config => $pi_config,
-			cgi => $cgi,
-			top => $top,
-		}) ]
-	];
-}
-
-# /$LISTNAME/?r=$GIT_COMMIT                 -> HTML only
-sub get_index {
-	my ($ctx, $cgi, $top) = @_;
+	PublicInbox::Feed::generate($ctx);
+}
+
+# /$INBOX/new.html			-> HTML only
+sub get_new {
+	my ($ctx) = @_;
 	require PublicInbox::Feed;
-	[ 200, [ 'Content-Type' => 'text/html; charset=UTF-8' ],
-	  [ PublicInbox::Feed->generate_html_index({
-			git_dir => $ctx->{git_dir},
-			listname => $ctx->{listname},
-			pi_config => $pi_config,
-			cgi => $cgi,
-			footer => footer($ctx),
-			top => $top,
-		}) ]
-	];
-}
-
-# just returns a string ref for the blob in the current ctx
-sub mid2blob {
+	PublicInbox::Feed::new_html($ctx);
+}
+
+# /$INBOX/?r=$GIT_COMMIT                 -> HTML only
+sub get_index {
 	my ($ctx) = @_;
-	require Digest::SHA;
-	my $hex = Digest::SHA::sha1_hex($ctx->{mid});
-	$hex =~ /\A([a-f0-9]{2})([a-f0-9]{38})\z/i or
-			die "BUG: not a SHA-1 hex: $hex";
-
-	my @cmd = ('git', "--git-dir=$ctx->{git_dir}",
-			qw(cat-file blob), "HEAD:$1/$2");
-	my $cmd = join(' ', @cmd);
-	my $pid = open my $fh, '-|';
-	defined $pid or die "fork failed: $!\n";
-	if ($pid == 0) {
-		open STDERR, '>', '/dev/null'; # ignore errors
-		exec @cmd or die "exec failed: $!\n";
+	require PublicInbox::Feed;
+	if ($ctx->{env}->{QUERY_STRING} =~ /(?:\A|[&;])q=/) {
+		require PublicInbox::SearchView;
+		PublicInbox::SearchView::sres_top_html($ctx);
 	} else {
-		my $blob = eval { local $/; <$fh> };
-		close $fh;
-		$? == 0 ? \$blob : undef;
+		PublicInbox::Feed::generate_html_index($ctx);
 	}
 }
 
-# /$LISTNAME/m/$MESSAGE_ID.txt                    -> raw original
+# /$INBOX/$MESSAGE_ID/raw                    -> raw mbox
 sub get_mid_txt {
-	my ($ctx, $cgi) = @_;
-	my $x = mid2blob($ctx);
-	$x ? [ 200, [ 'Content-Type' => 'text/plain' ], [ $$x ] ] : r404();
+	my ($ctx) = @_;
+	require PublicInbox::Mbox;
+	PublicInbox::Mbox::emit_raw($ctx) || r404($ctx);
 }
 
-# /$LISTNAME/m/$MESSAGE_ID.html                   -> HTML content (short quotes)
+# /$INBOX/$MESSAGE_ID/                   -> HTML content (short quotes)
 sub get_mid_html {
-	my ($ctx, $cgi) = @_;
-	my $x = mid2blob($ctx);
-	return r404() unless $x;
-
+	my ($ctx) = @_;
 	require PublicInbox::View;
-	my $mid_href = PublicInbox::Hval::ascii_html(
-						uri_escape_utf8($ctx->{mid}));
-	my $pfx = "../f/$mid_href.html";
-	my $foot = footer($ctx);
-	require Email::MIME;
-	[ 200, [ 'Content-Type' => 'text/html; charset=UTF-8' ],
-	  [ PublicInbox::View->msg_html(Email::MIME->new($x), $pfx, $foot) ] ];
-}
-
-# /$LISTNAME/f/$MESSAGE_ID.html                   -> HTML content (fullquotes)
-sub get_full_html {
-	my ($ctx, $cgi) = @_;
-	my $x = mid2blob($ctx);
-	return r404() unless $x;
+	PublicInbox::View::msg_page($ctx) || r404($ctx);
+}
+
+# /$INBOX/$MESSAGE_ID/t/
+sub get_thread {
+	my ($ctx, $flat) = @_;
+	$ctx->{-inbox}->over or return need($ctx, 'Overview');
+	$ctx->{flat} = $flat;
 	require PublicInbox::View;
-	require Email::MIME;
-	my $foot = footer($ctx);
-	[ 200, [ 'Content-Type' => 'text/html' ],
-	  [ PublicInbox::View->msg_html(Email::MIME->new($x), undef, $foot)] ];
+	PublicInbox::View::thread_html($ctx);
+}
+
+# /$INBOX/_/text/$KEY/
+# /$INBOX/_/text/$KEY/raw
+# KEY may contain slashes
+sub get_text {
+	my ($ctx, $inbox, $key) = @_;
+	my $r404 = invalid_inbox($ctx, $inbox);
+	return $r404 if $r404;
+
+	require PublicInbox::WwwText;
+	PublicInbox::WwwText::get_text($ctx, $key);
+}
+
+# show git objects (blobs and commits)
+# /$INBOX/_/$OBJECT_ID/show
+# /$INBOX/_/${OBJECT_ID}_${FILENAME}
+# KEY may contain slashes
+sub get_vcs_object ($$$;$) {
+	my ($ctx, $inbox, $oid, $filename) = @_;
+	my $r404 = invalid_inbox($ctx, $inbox);
+	return $r404 if $r404;
+	require PublicInbox::ViewVCS;
+	PublicInbox::ViewVCS::show($ctx, $oid, $filename);
+}
+
+sub need {
+	my ($ctx, $extra) = @_;
+	my $msg = <<EOF;
+<html><head><title>$extra not available for this
+public-inbox</title><body><pre>$extra is not available for this public-inbox
+<a href="../">Return to index</a></pre></body></html>
+EOF
+	[ 501, [ 'Content-Type' => 'text/html; charset=UTF-8' ], [ $msg ] ];
 }
 
-sub self_url {
-	my ($cgi) = @_;
-	ref($cgi) eq 'CGI' ? $cgi->self_url : $cgi->uri->as_string;
+# /$INBOX/$MESSAGE_ID/t.mbox           -> thread as mbox
+# /$INBOX/$MESSAGE_ID/t.mbox.gz        -> thread as gzipped mbox
+# note: I'm not a big fan of other compression formats since they're
+# significantly more expensive on CPU than gzip and less-widely available,
+# especially on older systems.  Stick to zlib since that's what git uses.
+sub get_thread_mbox {
+	my ($ctx, $sfx) = @_;
+	my $over = $ctx->{-inbox}->over or return need($ctx, 'Overview');
+	require PublicInbox::Mbox;
+	PublicInbox::Mbox::thread_mbox($ctx, $over, $sfx);
 }
 
-sub redirect_list_index {
-	my ($ctx, $cgi) = @_;
-	do_redirect(self_url($cgi) . "/");
+
+# /$INBOX/$MESSAGE_ID/t.atom		  -> thread as Atom feed
+sub get_thread_atom {
+	my ($ctx) = @_;
+	$ctx->{-inbox}->over or return need($ctx, 'Overview');
+	require PublicInbox::Feed;
+	PublicInbox::Feed::generate_thread_atom($ctx);
 }
 
-sub redirect_mid {
-	my ($ctx, $cgi) = @_;
-	my $url = self_url($cgi);
-	$url =~ s!/f/!/m/!;
-	do_redirect($url . '.html');
+sub legacy_redirects {
+	my ($ctx, $path_info) = @_;
+
+	# single-message pages
+	if ($path_info =~ m!$INBOX_RE/m/(\S+)/\z!o) {
+		r301($ctx, $1, $2);
+	} elsif ($path_info =~ m!$INBOX_RE/m/(\S+)/raw\z!o) {
+		r301($ctx, $1, $2, 'raw');
+
+	} elsif ($path_info =~ m!$INBOX_RE/f/(\S+)/\z!o) {
+		r301($ctx, $1, $2);
+
+	# thread display
+	} elsif ($path_info =~ m!$INBOX_RE/t/(\S+)/\z!o) {
+		r301($ctx, $1, $2, 't/#u');
+
+	} elsif ($path_info =~ m!$INBOX_RE/t/(\S+)/mbox(\.gz)?\z!o) {
+		r301($ctx, $1, $2, "t.mbox$3");
+
+	# even older legacy redirects
+	} elsif ($path_info =~ m!$INBOX_RE/m/(\S+)\.html\z!o) {
+		r301($ctx, $1, $2);
+
+	} elsif ($path_info =~ m!$INBOX_RE/t/(\S+)\.html\z!o) {
+		r301($ctx, $1, $2, 't/#u');
+
+	} elsif ($path_info =~ m!$INBOX_RE/f/(\S+)\.html\z!o) {
+		r301($ctx, $1, $2);
+
+	} elsif ($path_info =~ m!$INBOX_RE/(?:m|f)/(\S+)\.txt\z!o) {
+		r301($ctx, $1, $2, 'raw');
+
+	} elsif ($path_info =~ m!$INBOX_RE/t/(\S+)(\.mbox(?:\.gz)?)\z!o) {
+		r301($ctx, $1, $2, "t$3");
+
+	# legacy convenience redirects, order still matters
+	} elsif ($path_info =~ m!$INBOX_RE/m/(\S+)\z!o) {
+		r301($ctx, $1, $2);
+	} elsif ($path_info =~ m!$INBOX_RE/t/(\S+)\z!o) {
+		r301($ctx, $1, $2, 't/#u');
+	} elsif ($path_info =~ m!$INBOX_RE/f/(\S+)\z!o) {
+		r301($ctx, $1, $2);
+
+	# some Message-IDs have slashes in them and the HTTP server
+	# may try to be clever and unescape them :<
+	} elsif ($path_info =~ m!$INBOX_RE/(\S+/\S+)/$END_RE\z!o) {
+		msg_page($ctx, $1, $2, $3);
+
+	# in case people leave off the trailing slash:
+	} elsif ($path_info =~ m!$INBOX_RE/(\S+/\S+)/(T|t)\z!o) {
+		r301($ctx, $1, $2, $3 eq 't' ? 't/#u' : $3);
+	} elsif ($path_info =~ m!$INBOX_RE/(\S+/\S+)/f\z!o) {
+		r301($ctx, $1, $2);
+	} else {
+		news_cgit_fallback($ctx);
+	}
 }
 
-sub do_redirect {
-	my ($url) = @_;
+sub r301 {
+	my ($ctx, $inbox, $mid_ue, $suffix) = @_;
+	my $ibx = $ctx->{-inbox};
+	unless ($ibx) {
+		my $r404 = invalid_inbox($ctx, $inbox);
+		return $r404 if $r404;
+		$ibx = $ctx->{-inbox};
+	}
+	my $url = $ibx->base_url($ctx->{env});
+	my $qs = $ctx->{env}->{QUERY_STRING};
+	if (defined $mid_ue) {
+		# common, and much nicer as '@' than '%40':
+		$mid_ue =~ s/%40/@/g;
+		$url .= $mid_ue . '/';
+	}
+	$url .= $suffix if (defined $suffix);
+	$url .= "?$qs" if $qs ne '';
+
 	[ 301,
 	  [ Location => $url, 'Content-Type' => 'text/plain' ],
-	  [ "Redirecting to $url\n" ]
-	]
+	  [ "Redirecting to $url\n" ] ]
 }
 
-sub ctx_get {
-	my ($ctx, $key) = @_;
-	my $val = $ctx->{$key};
-	(defined $val && length $val) or die "BUG: bad ctx, $key unusable\n";
-	$val;
+sub msg_page {
+	my ($ctx, $inbox, $mid_ue, $e) = @_;
+	my $ret;
+	$ret = invalid_inbox_mid($ctx, $inbox, $mid_ue) and return $ret;
+	'' eq $e and return get_mid_html($ctx);
+	'T/' eq $e and return get_thread($ctx, 1);
+	't/' eq $e and return get_thread($ctx);
+	't.atom' eq $e and return get_thread_atom($ctx);
+	't.mbox' eq $e and return get_thread_mbox($ctx);
+	't.mbox.gz' eq $e and return get_thread_mbox($ctx, '.gz');
+	'raw' eq $e and return get_mid_txt($ctx);
+
+	# legacy, but no redirect for compatibility:
+	'f/' eq $e and return get_mid_html($ctx);
+	r404($ctx);
 }
 
-sub try_cat {
-	my ($path) = @_;
-	my $rv;
-	if (open(my $fh, '<', $path)) {
-		local $/;
-		$rv = <$fh>;
-		close $fh;
-	}
-	$rv;
+sub serve_git {
+	my ($ctx, $epoch, $path) = @_;
+	my $env = $ctx->{env};
+	my $ibx = $ctx->{-inbox};
+	my $git = defined $epoch ? $ibx->git_epoch($epoch) : $ibx->git;
+	$git ? PublicInbox::GitHTTPBackend::serve($env, $git, $path) : r404();
 }
 
-sub footer {
+sub mbox_results {
 	my ($ctx) = @_;
-	return '' unless $ctx;
-	my $git_dir = ctx_get($ctx, 'git_dir');
-
-	# favor user-supplied footer
-	my $footer = try_cat("$git_dir/public-inbox/footer.html");
-	if (defined $footer) {
-		chomp $footer;
-		return $footer;
+	if ($ctx->{env}->{QUERY_STRING} =~ /(?:\A|[&;])q=/) {
+		$ctx->{-inbox}->search or return need($ctx, 'search');
+		require PublicInbox::SearchView;
+		return PublicInbox::SearchView::mbox_results($ctx);
+	}
+	r404();
+}
+
+sub serve_mbox_range {
+	my ($ctx, $inbox, $range) = @_;
+	invalid_inbox($ctx, $inbox) || eval {
+		require PublicInbox::Mbox;
+		PublicInbox::Mbox::emit_range($ctx, $range);
+	}
+}
+
+sub news_www {
+	my ($self) = @_;
+	$self->{news_www} ||= do {
+		require PublicInbox::NewsWWW;
+		PublicInbox::NewsWWW->new($self->{pi_config});
 	}
+}
+
+sub cgit {
+	my ($self) = @_;
+	$self->{cgit} ||= do {
+		my $pi_config = $self->{pi_config};
+
+		if (defined($pi_config->{'publicinbox.cgitrc'})) {
+			require PublicInbox::Cgit;
+			PublicInbox::Cgit->new($pi_config);
+		} else {
+			require Plack::Util;
+			Plack::Util::inline_object(call => sub { r404() });
+		}
+	}
+}
+
+sub www_listing {
+	my ($self) = @_;
+	$self->{www_listing} ||= do {
+		require PublicInbox::WwwListing;
+		PublicInbox::WwwListing->new($self);
+	}
+}
 
-	# auto-generate a footer
-	my $listname = ctx_get($ctx, 'listname');
-	my $desc = try_cat("$git_dir/description");
-	$desc = '$GIT_DIR/description missing' unless defined $desc;
-	chomp $desc;
-
-	my $urls = try_cat("$git_dir/cloneurl");
-	my @urls = split(/\r?\n/, $urls || '');
-	my $nurls = scalar @urls;
-	if ($nurls == 0) {
-		$urls = '($GIT_DIR/cloneurl missing)';
-	} elsif ($nurls == 1) {
-		$urls = 'git archive URL for <a href="' . SSOMA_URL .
-			'">ssoma</a>: ' . $urls[0];
+# GET $INBOX/manifest.js.gz
+sub get_inbox_manifest ($$$) {
+	my ($ctx, $inbox, $key) = @_;
+	my $r404 = invalid_inbox($ctx, $inbox);
+	return $r404 if $r404;
+	require PublicInbox::WwwListing;
+	PublicInbox::WwwListing::js($ctx->{env}, [$ctx->{-inbox}]);
+}
+
+sub get_attach {
+	my ($ctx, $idx, $fn) = @_;
+	require PublicInbox::WwwAttach;
+	PublicInbox::WwwAttach::get_attach($ctx, $idx, $fn);
+}
+
+# User-generated content (UGC) may have excessively long lines
+# and screw up rendering on some browsers, so we use pre-wrap.
+#
+# We also force everything to the same scaled font-size because GUI
+# browsers (tested both Firefox and surf (webkit)) uses a larger font
+# for the Search <form> element than the rest of the page.  Font size
+# uniformity is important to people who rely on gigantic fonts.
+# Finally, we use monospace to ensure the Search field and button
+# has the same size and spacing as everything else which is
+# <pre>-formatted anyways.
+our $STYLE = 'pre{white-space:pre-wrap}*{font-size:100%;font-family:monospace}';
+
+sub stylesheets_prepare ($$) {
+	my ($self, $upfx) = @_;
+	my $mini = eval {
+		require CSS::Minifier;
+		sub { CSS::Minifier::minify(input => $_[0]) };
+	} || eval {
+		require CSS::Minifier::XS;
+		sub { CSS::Minifier::XS::minify($_[0]) };
+	} || sub { $_[0] };
+
+	my $css_map = {};
+	my $stylesheets = $self->{pi_config}->{css} || [];
+	my $links = [];
+	my $inline_ok = 1;
+
+	foreach my $s (@$stylesheets) {
+		my $attr = {};
+		local $_ = $s;
+		foreach my $k (qw(media title href)) {
+			if (s/\s*$k='([^']+)'// || s/\s*$k=(\S+)//) {
+				$attr->{$k} = $1;
+			}
+		}
+
+		if (defined $attr->{href}) {
+			$inline_ok = 0;
+		} else {
+			my $fn = $_;
+			my ($key) = (m!([^/]+?)(?:\.css)?\z!i);
+			if ($key !~ /\A[a-zA-Z0-9_\-\.]+\z/) {
+				warn "ignoring $fn, non-ASCII word character\n";
+				next;
+			}
+			open(my $fh, '<', $fn) or do {
+				warn "failed to open $fn: $!\n";
+				next;
+			};
+			my $ctime = 0;
+			my $local = do { local $/; <$fh> };
+			if ($local =~ /\S/) {
+				$ctime = sprintf('%x',(stat($fh))[10]);
+				$local = $mini->($local);
+			}
+
+			# do not let BOFHs override userContent.css:
+			if ($local =~ /!\s*important\b/i) {
+				warn "ignoring $fn since it uses `!important'\n";
+				next;
+			}
+
+			$css_map->{$key} = $local;
+			$attr->{href} = "$upfx$key.css?$ctime";
+			if (defined($attr->{title})) {
+				$inline_ok = 0;
+			} elsif (($attr->{media}||'screen') eq 'screen') {
+				$attr->{-inline} = $local;
+			}
+		}
+		push @$links, $attr;
+	}
+
+	my $buf = "<style>$STYLE";
+	if ($inline_ok) {
+		my @ext; # for media=print and whatnot
+		foreach my $attr (@$links) {
+			if (defined(my $str = delete $attr->{-inline})) {
+				$buf .= $str;
+			} else {
+				push @ext, $attr;
+			}
+		}
+		$links = \@ext;
+	}
+	$buf .= '</style>';
+
+	if (@$links) {
+		foreach my $attr (@$links) {
+			delete $attr->{-inline};
+			$buf .= "<link\ntype=text/css\nrel=stylesheet";
+			while (my ($k, $v) = each %$attr) {
+				$v = qq{"$v"} if $v =~ /[\s=]/;
+				$buf .= qq{\n$k=$v};
+			}
+			$buf .= ' />';
+		}
+		$self->{"-style-$upfx"} = $buf;
 	} else {
-		$urls = 'git archive URLs for <a href="' . SSOMA_URL .
-			"\">ssoma</a>:\n" . join('', map { "\t" . $_ } @urls);
+		$self->{-style_inline} = $buf;
 	}
+	$self->{-css_map} = $css_map;
+}
+
+# returns an HTML fragment with <style> or <link> tags in them
+# Called by WwwStream by nearly every HTML page
+sub style {
+	my ($self, $upfx) = @_;
+	$self->{-style_inline} || $self->{"-style-$upfx"} || do {
+		stylesheets_prepare($self, $upfx);
+		$self->{-style_inline} || $self->{"-style-$upfx"}
+	};
+}
 
-	my $addr = $pi_config->get($listname, 'address');
-	if (ref($addr) eq 'ARRAY') {
-		$addr = $addr->[0]; # first address is primary
+# /$INBOX/$KEY.css endpoint
+# CSS is configured globally for all inboxes, but we access them on
+# a per-inbox basis.  This allows administrators to setup per-inbox
+# static routes to intercept the request before it hits PSGI
+sub get_css ($$$) {
+	my ($ctx, $inbox, $key) = @_;
+	my $r404 = invalid_inbox($ctx, $inbox);
+	return $r404 if $r404;
+	my $self = $ctx->{www};
+	my $css_map = $self->{-css_map} || stylesheets_prepare($self, '');
+	my $css = $css_map->{$key};
+	if (!defined($css) && $key eq 'userContent') {
+		my $env = $ctx->{env};
+		$css = PublicInbox::UserContent::sample($ctx->{-inbox}, $env);
 	}
+	defined $css or return r404();
+	my $h = [ 'Content-Length', bytes::length($css),
+		'Content-Type', 'text/css' ];
+	PublicInbox::GitHTTPBackend::cache_one_year($h);
+	[ 200, $h, [ $css ] ];
+}
 
-	$addr = "<a href=\"mailto:$addr\">$addr</a>";
-	$desc =  $desc;
-	join("\n",
-		'- ' . $desc,
-		'This is a <a href="' . PI_URL . '">public-inbox</a>, '.
-		"anybody may post:",
-		"\t$addr (text-only, no HTML please)",
-		$urls
-	);
+sub get_description {
+	my ($ctx, $inbox) = @_;
+	invalid_inbox($ctx, $inbox) || do {
+		my $d = $ctx->{-inbox}->description . "\n";
+		[ 200, [ 'Content-Length', bytes::length($d),
+			'Content-Type', 'text/plain' ], [ $d ] ];
+	};
 }
 
 1;