release: 0.6.0: Delivery detail page, DNS badge, machine metrics, wide layout

A page per delivery (`/deliveries/{id}`), reached from the *Details* link on
Added:
  - A page per delivery (`/deliveries/{id}`), reached from the *Details* link on
  - A **DNS** badge in the domain list, one per row, carrying the same
  - Machine metrics on the status page: a **Machine** card reporting the
  - An index of the current page's own sections in the navigation column, for the
Changed:
  - Every panel page is laid out in one column of the same width, so moving
  - The mark's small-size variant — the tab icon and the `SP` initials it
  - The stamp's `SELF-HOSTED SMTP RELAY` line is set at 11.5/0.15 instead of
  - The delivery log lists what identifies a message and nothing else: time,
  - Subjects are decoded for display as well as on the way in, so the rows the
  … (4 more in CHANGELOG)

Full detail: CHANGELOG.md (0.6.0).

internal/mailhdr/subject.go

@@ -0,0 +1,38 @@
+// Package mailhdr turns raw mail header values into display text. It sits
+// between the journal-milter, which reads headers off the wire, and the panel,
+// which shows them: both need the same answer for the same header, and the
+// panel needs it for rows the milter wrote before this decoding existed.
+package mailhdr
+
+import (
+	"mime"
+	"strings"
+)
+
+// SubjectMaxRunes caps what the journal keeps of a subject. A Subject header
+// may legally run to hundreds of characters; the log only needs enough to
+// recognise the message, and the panel shows one row per recipient.
+const SubjectMaxRunes = 200
+
+// DecodeSubject turns a raw Subject header into display text. Anything
+// non-ASCII arrives as RFC 2047 encoded-words (=?utf-8?Q?=D0=9F…?=), which the
+// panel would otherwise show verbatim: unreadable, and — being one unbreakable
+// run — wide enough to push the send-log table out of its card. Go's decoder
+// covers the UTF-8 and ASCII charsets senders use in practice; for anything
+// else (windows-1251, koi8-r) it fails and the raw header is kept, which is no
+// worse than not decoding at all. Truncation is applied after decoding so the
+// cap counts characters of the subject, not bytes of its encoding.
+//
+// It is idempotent: already-decoded text contains no encoded-words, so a second
+// pass returns it unchanged. That is what lets the panel decode on the way out
+// as well as the milter on the way in.
+func DecodeSubject(v string) string {
+	if dec, err := (&mime.WordDecoder{}).DecodeHeader(v); err == nil {
+		v = dec
+	}
+	v = strings.TrimSpace(v)
+	if r := []rune(v); len(r) > SubjectMaxRunes {
+		v = string(r[:SubjectMaxRunes]) + "…"
+	}
+	return v
+}