diff --git a/README.md b/README.md
index 992ac72..e2c2693 100644
--- a/README.md
+++ b/README.md
@@ -144,16 +144,16 @@ Edits show instantly and save in the background.
 
 ## Importing a list
 
-**Import** in the list editor adds entries from a file, after the ones already there and in the file's order:
+**Import** in the list editor adds entries from a file. **Line numbers are places**: line 1 becomes #1, line 38 becomes #38, and a blank line leaves its place empty (entries already at those places move down to make room).
 
-- **Text** (`.txt`, or paste): one title per line. Numbering like `1.` is dropped when every line has it, and a trailing `(1999)` is read as the year. For songs and albums, `Artist - Title` works either way round.
-- **CSV / TSV**: a header row picks the columns: `Title` (or Name), `Year`, `Artist`/`Director`/`Developer`, `Note` (or Why), and `Rank` to set the order. Without a header the first column is the title. Canon's own export imports as-is.
+- **Text** (`.txt`, or paste): one title per line. If every line is numbered (`38. Super Mario World`), the numbers are the places instead. A trailing `(1999)` is read as the year. For songs and albums, `Artist - Title` works either way round.
+- **CSV / TSV**: a header row picks the columns: `Title` (or Name), `Year`, `Artist`/`Director`/`Developer`, `Note` (or Why), and `Rank` to give places directly. Without a header the first column is the title. The first row after the header is #1.
 
 Each entry is searched for and matched. Canon only adds a match on its own when it's certain: the title matches exactly and nothing else fits as well (a year or artist in the file helps). Everything else is left for you to check, with a recommended pick, the other results, **Add as written** to make a new entry from your text, **Skip**, and a box to search again. Custom lists search all of Wikipedia, so there a match is only certain when a year or subtitle confirms it.
 
 ## Exporting a list
 
-Every list can be downloaded as its titles only, in rank order, one per line. There are no ranks, header row or other fields.
+Every list can be downloaded as its titles only, one per line, where line N is place #N: an empty place is a blank line. There are no ranks, header row or other fields, and importing the file puts everything back where it was.
 
 - **Export → Titles as CSV** (`/l/<list>/export?format=csv`): a single column. It starts with a UTF-8 BOM so Excel shows accented titles correctly.
 - **Export → Titles as text** (`/l/<list>/export?format=txt`): plain lines.
diff --git a/internal/importer/importer_test.go b/internal/importer/importer_test.go
index 98f4999..5c73f2a 100644
--- a/internal/importer/importer_test.go
+++ b/internal/importer/importer_test.go
@@ -1,6 +1,7 @@
 package importer
 
 import (
+	"fmt"
 	"reflect"
 	"testing"
 
@@ -8,17 +9,28 @@ import (
 )
 
 func TestParseText(t *testing.T) {
-	got, err := Parse([]byte("1. Heat (1995)\n2) 2001: A Space Odyssey\n\n3 - Alien [1979]\n"), "list.txt")
+	got, err := Parse([]byte("1. Heat (1995)\n2) 2001: A Space Odyssey\n\n4 - Alien [1979]\n"), "list.txt")
 	if err != nil {
 		t.Fatal(err)
 	}
-	want := []Row{{Title: "Heat", Year: "1995"}, {Title: "2001: A Space Odyssey"}, {Title: "Alien", Year: "1979"}}
+	want := []Row{{Place: 1, Title: "Heat", Year: "1995"}, {Place: 2, Title: "2001: A Space Odyssey"}, {Place: 4, Title: "Alien", Year: "1979"}}
 	if !reflect.DeepEqual(got, want) {
 		t.Fatalf("got %+v", got)
 	}
-	// Not every line is numbered, so nothing is stripped.
-	got, _ = Parse([]byte("2001: A Space Odyssey\nHeat\n1917\n"), "x.txt")
-	if got[0].Title != "2001: A Space Odyssey" || got[2].Title != "1917" {
+	// Not every line is numbered, so nothing is stripped; line numbers are
+	// the places, and a blank line leaves one empty.
+	got, _ = Parse([]byte("2001: A Space Odyssey\nHeat\n\n1917\n"), "x.txt")
+	if got[0].Title != "2001: A Space Odyssey" || got[2].Title != "1917" || got[2].Place != 4 {
+		t.Fatalf("got %+v", got)
+	}
+	// Numbers win over line numbers when every line has one.
+	got, _ = Parse([]byte("1. Super Mario Sunshine\n38. Super Mario World\n"), "games.txt")
+	if got[0].Place != 1 || got[1].Place != 38 || got[1].Title != "Super Mario World" {
+		t.Fatalf("got %+v", got)
+	}
+	// A repeated number takes the next free place, keeping the file's order.
+	got, _ = Parse([]byte("3. A\n3. B\n1. C\n"), "x.txt")
+	if fmt.Sprint(got) != fmt.Sprint([]Row{{Place: 1, Title: "C"}, {Place: 3, Title: "A"}, {Place: 4, Title: "B"}}) {
 		t.Fatalf("got %+v", got)
 	}
 }
@@ -30,20 +42,26 @@ func TestParseCSV(t *testing.T) {
 		t.Fatal(err)
 	}
 	want := []Row{
-		{Title: "Heat", Year: "1995", Subtitle: "Michael Mann"},
-		{Title: "Alien", Year: "1979", Subtitle: "Ridley Scott", Note: "Tense, perfect"},
+		{Place: 1, Title: "Heat", Year: "1995", Subtitle: "Michael Mann"},
+		{Place: 2, Title: "Alien", Year: "1979", Subtitle: "Ridley Scott", Note: "Tense, perfect"},
 	}
 	if !reflect.DeepEqual(got, want) {
 		t.Fatalf("got %+v", got)
 	}
-	// Canon's own export: one quoted column, no header.
-	got, _ = Parse([]byte("\xef\xbb\xbf\"Heat\"\r\n\"Crouching Tiger, Hidden Dragon\"\r\n"), "films.csv")
-	if len(got) != 2 || got[1].Title != "Crouching Tiger, Hidden Dragon" {
+	// Canon's own export: one column, no header, a blank line for an empty
+	// place.
+	got, _ = Parse([]byte("\xef\xbb\xbfHeat\r\n\r\n\"Crouching Tiger, Hidden Dragon\"\r\n"), "films.csv")
+	if len(got) != 2 || got[1].Title != "Crouching Tiger, Hidden Dragon" || got[1].Place != 3 {
+		t.Fatalf("got %+v", got)
+	}
+	// With a header, the first row after it is #1.
+	got, _ = Parse([]byte("Title\nHeat\n\nAlien\n"), "x.csv")
+	if got[0].Place != 1 || got[1].Place != 3 {
 		t.Fatalf("got %+v", got)
 	}
 	// Headerless rank,title,year.
-	got, _ = Parse([]byte("1,Heat,1995\n2,Alien,1979\n"), "x.csv")
-	if got[1] != (Row{Title: "Alien", Year: "1979"}) {
+	got, _ = Parse([]byte("1,Heat,1995\n5,Alien,1979\n"), "x.csv")
+	if got[1] != (Row{Place: 5, Title: "Alien", Year: "1979"}) {
 		t.Fatalf("got %+v", got)
 	}
 	if _, err := Parse([]byte("\n \n"), "x.txt"); err != ErrEmpty {
diff --git a/internal/importer/parse.go b/internal/importer/parse.go
index be12b00..5eb1b9c 100644
--- a/internal/importer/parse.go
+++ b/internal/importer/parse.go
@@ -6,6 +6,7 @@ import (
 	"bytes"
 	"encoding/csv"
 	"errors"
+	"io"
 	"regexp"
 	"slices"
 	"strconv"
@@ -16,8 +17,12 @@ import (
 // MaxRows caps one import.
 const MaxRows = 2000
 
-// Row is one entry read from the file, in the file's order.
+// Row is one entry read from the file.
 type Row struct {
+	// Place is where it goes on the list: its line number (blank lines
+	// leave a place empty), or the number it was given ("38. Title" in a
+	// numbered list, or a Rank column).
+	Place    int    `json:"place"`
 	Title    string `json:"title"`
 	Year     string `json:"year,omitempty"`
 	Subtitle string `json:"subtitle,omitempty"` // artist, director, developer…
@@ -50,12 +55,27 @@ func Parse(data []byte, filename string) ([]Row, error) {
 	if len(rows) == 0 {
 		return nil, ErrEmpty
 	}
+	rows = orderPlaces(rows)
 	if len(rows) > MaxRows {
 		rows = rows[:MaxRows]
 	}
 	return rows, nil
 }
 
+// orderPlaces sorts rows by place and gives a repeated place (two rows both
+// numbered 3) the next free one, keeping the file's order.
+func orderPlaces(rows []Row) []Row {
+	slices.SortStableFunc(rows, func(a, b Row) int { return a.Place - b.Place })
+	prev := 0
+	for i := range rows {
+		if rows[i].Place <= prev {
+			rows[i].Place = prev + 1
+		}
+		prev = rows[i].Place
+	}
+	return rows
+}
+
 // latin1 decodes bytes as Windows-1252/Latin-1, which is what spreadsheet
 // exports that aren't UTF-8 usually are.
 func latin1(b []byte) string {
@@ -84,27 +104,35 @@ var numbered = regexp.MustCompile(`^\s*#?\d{1,5}\s*[.):\-–—]?\s+`)
 var trailingYear = regexp.MustCompile(`\s*[(\[]((?:1[5-9]|20)\d\d)[)\]]\s*$`)
 
 func parseLines(text string) []Row {
-	var lines []string
-	for _, l := range strings.Split(text, "\n") {
+	lines := strings.Split(text, "\n")
+	// Only use numbering when every non-blank line has it, so a list that
+	// merely contains "2001: A Space Odyssey" keeps its title.
+	numberedAll, any := true, false
+	for _, l := range lines {
 		if l = strings.TrimSpace(l); l != "" {
-			lines = append(lines, l)
+			any = true
+			if !numbered.MatchString(l) {
+				numberedAll = false
+				break
+			}
 		}
 	}
-	// Only strip numbering when every line has it, so a list that merely
-	// contains "2001: A Space Odyssey" keeps its title.
-	strip := len(lines) > 0
-	for _, l := range lines {
-		if !numbered.MatchString(l) {
-			strip = false
-			break
+	numberedAll = numberedAll && any
+	var rows []Row
+	for i, l := range lines {
+		l = strings.TrimSpace(l)
+		if l == "" {
+			continue // a blank line is an empty place
 		}
-	}
-	rows := make([]Row, 0, len(lines))
-	for _, l := range lines {
-		if strip {
+		place := i + 1
+		if numberedAll {
+			if n, err := strconv.Atoi(strings.Trim(numbered.FindString(l), " \t#.):-–—")); err == nil && n > 0 {
+				place = n
+			}
 			l = numbered.ReplaceAllString(l, "")
 		}
 		if r := rowFromTitle(l); r.Title != "" {
+			r.Place = place
 			rows = append(rows, r)
 		}
 	}
@@ -140,18 +168,28 @@ func parseDelimited(text string, delim rune) ([]Row, error) {
 	cr.FieldsPerRecord = -1
 	cr.LazyQuotes = true
 	cr.TrimLeadingSpace = true
-	records, err := cr.ReadAll()
-	if err != nil {
-		return nil, errors.New("that CSV file couldn't be read: " + err.Error())
+	// The reader skips blank lines, so keep each record's line number: a
+	// blank line is an empty place.
+	var records [][]string
+	var lines []int
+	for {
+		rec, err := cr.Read()
+		if err == io.EOF {
+			break
+		}
+		if err != nil {
+			return nil, errors.New("that CSV file couldn't be read: " + err.Error())
+		}
+		line, _ := cr.FieldPos(0)
+		records = append(records, rec)
+		lines = append(lines, line)
 	}
-	records = slices.DeleteFunc(records, func(rec []string) bool {
-		return strings.TrimSpace(strings.Join(rec, "")) == ""
-	})
 	if len(records) == 0 {
 		return nil, nil
 	}
 
 	cols := map[string]int{}
+	first := 1 // the line place #1 is on
 	if head := records[0]; isHeader(head) {
 		for i, name := range head {
 			if role, ok := headerNames[strings.ToLower(strings.TrimSpace(name))]; ok {
@@ -160,7 +198,8 @@ func parseDelimited(text string, delim rune) ([]Row, error) {
 				}
 			}
 		}
-		records = records[1:]
+		first = lines[0] + 1
+		records, lines = records[1:], lines[1:]
 	} else {
 		cols = guessColumns(records)
 	}
@@ -168,11 +207,6 @@ func parseDelimited(text string, delim rune) ([]Row, error) {
 		cols["title"] = 0
 	}
 
-	type ranked struct {
-		Row
-		rank int
-	}
-	var out []ranked
 	cell := func(rec []string, role string) string {
 		i, ok := cols[role]
 		if !ok || i >= len(rec) {
@@ -180,6 +214,7 @@ func parseDelimited(text string, delim rune) ([]Row, error) {
 		}
 		return strings.TrimSpace(rec[i])
 	}
+	var rows []Row
 	for i, rec := range records {
 		r := Row{Title: cell(rec, "title"), Subtitle: cell(rec, "subtitle"), Note: cell(rec, "note")}
 		if y := yearIn(cell(rec, "year")); y != "" {
@@ -191,18 +226,11 @@ func parseDelimited(text string, delim rune) ([]Row, error) {
 		if r.Title == "" {
 			continue
 		}
-		rank, err := strconv.Atoi(strings.TrimPrefix(cell(rec, "rank"), "#"))
-		if err != nil {
-			rank = 1 << 30
+		r.Place = lines[i] - first + 1
+		if n, err := strconv.Atoi(strings.TrimPrefix(cell(rec, "rank"), "#")); err == nil && n > 0 {
+			r.Place = n
 		}
-		out = append(out, ranked{r, rank*MaxRows*2 + i}) // stable within equal ranks
-	}
-	if _, ok := cols["rank"]; ok {
-		slices.SortStableFunc(out, func(a, b ranked) int { return a.rank - b.rank })
-	}
-	rows := make([]Row, len(out))
-	for i, r := range out {
-		rows[i] = r.Row
+		rows = append(rows, r)
 	}
 	return rows, nil
 }
diff --git a/internal/server/export.go b/internal/server/export.go
index a29e4be..e0ca027 100644
--- a/internal/server/export.go
+++ b/internal/server/export.go
@@ -14,8 +14,9 @@ var exportFormats = map[string]struct{ ext, contentType string }{
 	"txt": {"txt", "text/plain; charset=utf-8"},
 }
 
-// handleExport downloads a list's titles in rank order, one per line, with
-// nothing else: no ranks, header row or metadata.
+// handleExport downloads a list's titles, one per line, with nothing else:
+// no ranks, header row or metadata. Line N is place #N, so an empty place
+// is a blank line, and importing the file puts everything back where it was.
 func (s *Server) handleExport(w http.ResponseWriter, r *http.Request) {
 	format := r.URL.Query().Get("format")
 	if format == "" {
@@ -35,9 +36,12 @@ func (s *Server) handleExport(w http.ResponseWriter, r *http.Request) {
 		s.serverError(w, r, err)
 		return
 	}
-	titles := make([]string, len(items))
-	for i, it := range items {
-		titles[i] = it.Title
+	var titles []string
+	for _, it := range items {
+		for len(titles) < it.Rank-1 {
+			titles = append(titles, "")
+		}
+		titles = append(titles, it.Title)
 	}
 	w.Header().Set("Content-Type", f.contentType)
 	w.Header().Set("Content-Disposition", fmt.Sprintf(`attachment; filename="%s.%s"`, l.Slug, f.ext))
diff --git a/internal/server/server.go b/internal/server/server.go
index c66fb64..895e4ec 100644
--- a/internal/server/server.go
+++ b/internal/server/server.go
@@ -321,20 +321,45 @@ func (s *Server) isHTTPS(r *http.Request) bool {
 
 func (s *Server) staticHandler() http.Handler {
 	fsrv := http.FileServerFS(s.static)
+	etags := staticETags(s.static)
 	return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		if strings.HasSuffix(r.URL.Path, "/") {
 			s.notFound(w, r)
 			return
 		}
+		if tag, ok := etags[strings.TrimPrefix(r.URL.Path, "/")]; ok {
+			w.Header().Set("ETag", tag) // FileServer answers If-None-Match with 304
+		}
 		if r.URL.Query().Get("v") != "" {
 			w.Header().Set("Cache-Control", "public, max-age=31536000, immutable")
 		} else {
-			w.Header().Set("Cache-Control", "public, max-age=3600")
+			// Scripts load each other without a ?v= (import "./places.js"),
+			// so unversioned files are re-checked on every load; the ETag
+			// makes that a quick 304 until a new build changes them.
+			w.Header().Set("Cache-Control", "no-cache")
 		}
 		fsrv.ServeHTTP(w, r)
 	})
 }
 
+// staticETags fingerprints every embedded static file once at startup.
+func staticETags(fsys fs.FS) map[string]string {
+	tags := map[string]string{}
+	fs.WalkDir(fsys, ".", func(p string, d fs.DirEntry, err error) error {
+		if err != nil || d.IsDir() {
+			return nil
+		}
+		b, err := fs.ReadFile(fsys, p)
+		if err != nil {
+			return nil
+		}
+		sum := sha256.Sum256(b)
+		tags[p] = `"` + hex.EncodeToString(sum[:8]) + `"`
+		return nil
+	})
+	return tags
+}
+
 func (s *Server) handleImage(w http.ResponseWriter, r *http.Request) {
 	name := r.PathValue("name")
 	if !imgcache.ValidName(name) {
diff --git a/web/static/css/app.css b/web/static/css/app.css
index 2da24e5..2d1417e 100644
--- a/web/static/css/app.css
+++ b/web/static/css/app.css
@@ -889,12 +889,12 @@ input[type="checkbox"] { accent-color: var(--accent); width: 18px; height: 18px;
 .import-row { border: 1px solid var(--border); border-radius: var(--radius); padding: 10px 12px; display: flex; flex-direction: column; gap: 10px; background: var(--surface); }
 .import-row.is-review, .import-row.is-none, .import-row.is-error { border-color: color-mix(in srgb, var(--warn) 45%, var(--border)); }
 .import-row-head { display: flex; align-items: center; gap: 10px; min-width: 0; }
-.import-num { color: var(--faint); font-size: 13px; font-variant-numeric: tabular-nums; min-width: 2.5ch; text-align: right; }
+.import-num { color: var(--faint); font-size: 13px; font-variant-numeric: tabular-nums; min-width: 4.5ch; text-align: right; }
 .import-src { flex: 1; min-width: 0; font-size: 14px; overflow: hidden; text-overflow: ellipsis; white-space: nowrap; }
 .import-src b { font-weight: 600; }
 .badge--ok { background: color-mix(in srgb, var(--success) 16%, transparent); color: var(--success); }
 .badge .icon { width: 12px; height: 12px; }
-.import-chosen { display: flex; align-items: center; gap: 10px; justify-content: space-between; padding-left: calc(2.5ch + 10px); }
+.import-chosen { display: flex; align-items: center; gap: 10px; justify-content: space-between; padding-left: calc(4.5ch + 10px); }
 .import-card { display: flex; align-items: center; gap: 10px; min-width: 0; }
 .import-card .result-art { width: 38px; flex: none; }
 .import-card-text { display: flex; flex-direction: column; gap: 2px; min-width: 0; }
diff --git a/web/static/js/import.js b/web/static/js/import.js
index d7334a2..2145abc 100644
--- a/web/static/js/import.js
+++ b/web/static/js/import.js
@@ -32,7 +32,7 @@ export function openImport({ list, cat, onListChange, onDone }) {
     dialogHead("Import a list", () => dlg),
     h("div", { class: "dialog-body stack" },
       h("p", { class: "muted" },
-        `Entries are added after your last one, in the file's order. Use one title per line, or a CSV with columns like Title, Year, ${artistWord(cat)} and Note. `,
+        `Line 1 becomes #1, line 38 becomes #38, and a blank line leaves its place empty${list.itemCount ? " (entries already at those places move down to make room)" : ""}. Use one title per line, or a CSV with columns like Title, Year, ${artistWord(cat)} and Note. `,
         "Each one is looked up and matched for you; anything Canon isn't sure about is left for you to check."),
       drop,
       text,
@@ -117,14 +117,18 @@ export function openImport({ list, cat, onListChange, onDone }) {
         ].filter(Boolean);
       summary.textContent = parts.join(" · ");
 
-      const n = chosen().length;
-      const free = list.maxItems - list.itemCount;
-      go.disabled = matching || n === 0 || n > free;
+      const picked = chosen();
+      const n = picked.length;
+      const lastPlace = Math.max(0, ...picked.map((e) => e.row.place));
+      const fits = list.itemCount + n <= list.maxItems && lastPlace <= list.maxItems;
+      go.disabled = matching || n === 0 || !fits;
       go.textContent = matching ? "Matching…" : `Import ${fmt.format(n)} ${n === 1 ? "entry" : "entries"}`;
-      room.hidden = matching || n <= free;
+      room.hidden = matching || n === 0 || fits;
       if (!room.hidden) {
-        const want = Math.min(list.itemCount + n, 65535);
-        room.replaceChildren(icon("info"), `This list has room for ${fmt.format(free)} more. `,
+        const want = Math.min(Math.max(list.itemCount + n, lastPlace), 65535);
+        room.replaceChildren(icon("info"), lastPlace > list.maxItems
+          ? `The file goes up to #${fmt.format(lastPlace)}, but this list ends at #${fmt.format(list.maxItems)}. `
+          : `This list has room for ${fmt.format(list.maxItems - list.itemCount)} more. `,
           h("button", {
             type: "button", class: "btn btn--sm", onclick: async (e) => {
               busy(e.currentTarget, true);
@@ -163,7 +167,7 @@ export function openImport({ list, cat, onListChange, onDone }) {
 
       const li = h("li", { class: `import-row is-${e.state}`, "data-i": e.i },
         h("div", { class: "import-row-head" },
-          h("span", { class: "import-num" }, `${e.i + 1}.`),
+          h("span", { class: "import-num" }, `#${row.place}`),
           h("div", { class: "import-src" }, h("b", {}, row.title), asWritten && h("span", { class: "muted small" }, ` ${asWritten}`)),
           badge),
       );
@@ -299,9 +303,9 @@ export function openImport({ list, cat, onListChange, onDone }) {
         ? {
           title: r.title, subtitle: r.subtitle, year: r.year, externalUrl: r.externalUrl, imageUrl: r.imageUrl,
           provider: r.provider, providerId: r.providerId, extra: { ...(r.extra || {}) }, description: r.overview || "",
-          details: true, note: e.row.note || "",
+          details: true, note: e.row.note || "", rank: e.row.place,
         }
-        : { title: e.row.title, subtitle: e.row.subtitle || "", year: e.row.year || "", note: e.row.note || "", extra: {} };
+        : { title: e.row.title, subtitle: e.row.subtitle || "", year: e.row.year || "", note: e.row.note || "", extra: {}, rank: e.row.place };
       try {
         const res = await api("POST", `/api/lists/${list.id}/items`, payload);
         added++;