switch to gzipped mboxes

Mboxes may be huge, so only support downloading gzipped mboxes
to save bandwidth and to get free checksumming.

Streaming output means we should not be wasting too much memory
on this unless the chosen server sucks.

1 files changed, 6 insertions, 2 deletions

diff --git a/lib/PublicInbox/WWW.pm b/lib/PublicInbox/WWW.pm
index cd8a5705..68839d7c 100644
--- a/lib/PublicInbox/WWW.pm
+++ b/lib/PublicInbox/WWW.pm
@@ -53,7 +53,8 @@ sub run {
         } elsif ($path_info =~ m!$LISTNAME_RE/t/(\S+)\.html\z!o) {
                 invalid_list_mid(\%ctx, $1, $2) || get_thread(\%ctx, $cgi);
 
-        } elsif ($path_info =~ m!$LISTNAME_RE/t/(\S+)\.mbox\z!o) {
+        } elsif ($path_info =~ m!$LISTNAME_RE/t/(\S+)\.mbox\.gz!o) {
+                my $sfx = $3;
                 invalid_list_mid(\%ctx, $1, $2) || get_thread_mbox(\%ctx, $cgi);
 
         } elsif ($path_info =~ m!$LISTNAME_RE/f/\S+\.txt\z!o) {
@@ -329,7 +330,10 @@ sub msg_pfx {
         "../f/$href.html";
 }
 
-# /$LISTNAME/t/$MESSAGE_ID.mbox                    -> search results as mbox
+# /$LISTNAME/t/$MESSAGE_ID.mbox.gz        -> search results 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, $cgi) = @_;
         my $srch = searcher($ctx) or return need_search($ctx);