diff --git a/src/Init/Data/String/Basic.lean b/src/Init/Data/String/Basic.lean
index 1f2567cfdf..94cffd92df 100644
--- a/src/Init/Data/String/Basic.lean
+++ b/src/Init/Data/String/Basic.lean
@@ -12,8 +12,11 @@ universes u
 structure String :=
 (data : List Char)
 
-/-- A character position in an UTF-8 encoded string.
-To represent codepoint positions, use a plain `Nat`. -/
+/-- A byte position in a `String`. Internally, `String`s are UTF-8 encoded.
+Codepoint positions (counting the Unicode codepoints rather than bytes)
+are represented by plain `Nat`s instead.
+Indexing a `String` by a byte position is constant-time, while codepoint
+positions need to be translated internally to byte positions in linear-time. -/
 abbrev String.Pos := Nat
 
 structure Substring :=
diff --git a/src/Lean/Data/Lsp/Utf16.lean b/src/Lean/Data/Lsp/Utf16.lean
index d0d041760e..b4db02772f 100644
--- a/src/Lean/Data/Lsp/Utf16.lean
+++ b/src/Lean/Data/Lsp/Utf16.lean
@@ -72,12 +72,5 @@ text.source.codepointPosToUtf8PosFrom colPos chr
 def leanPosToLspPos (text : FileMap) : Lean.Position → Lsp.Position
 | ⟨ln, col⟩ => ⟨ln-1, text.source.codepointPosToUtf16PosFrom col (text.positions.get! $ ln - 1)⟩
 
-def replaceLspRange (text : FileMap) (r : Lsp.Range) (newText : String) : FileMap :=
-let start := text.lspPosToUtf8Pos r.start;
-let «end» := text.lspPosToUtf8Pos r.«end»;
-let pre := text.source.extract 0 start;
-let post := text.source.extract «end» text.source.bsize;
-(pre ++ newText ++ post).toFileMap
-
 end FileMap
 end Lean
diff --git a/src/Lean/Server.lean b/src/Lean/Server.lean
index 91421ebe36..c56fa1a2ce 100644
--- a/src/Lean/Server.lean
+++ b/src/Lean/Server.lean
@@ -130,17 +130,23 @@ writeLspNotification "textDocument/publishDiagnostics"
 
 def handleDidOpen (p : DidOpenTextDocumentParams) : ServerM Unit := do
 let doc := p.textDocument;
--- LSP is EOL agnostic, so we first split on "\r\n" to avoid possibly redundant line breaks.
-let splitOnEOLs (s : String) : List String := (do
-  line ← s.splitOn "\r\n";
-  line ← line.splitOn "\n";
-  line ← line.splitOn "\r";
-  pure line);
-let text := ("\n".intercalate $ splitOnEOLs doc.text).toFileMap;
+-- NOTE(WN): `toFileMap` marks line beginnings as immediately following
+-- "\n", which should be enough to handle both LF and CRLF correctly.
+-- This is because LSP always refers to characters by (line, column),
+-- so if we get the line number correct it shouldn't matter that there
+-- is a CR there.
+let text := doc.text.toFileMap;
 (msgLog, newDoc) ← monadLift $ compileDocument doc.version text;
 updateOpenDocuments doc.uri newDoc;
 sendDiagnostics doc.uri newDoc msgLog
 
+private def replaceLspRange (text : FileMap) (r : Lsp.Range) (newText : String) : FileMap :=
+let start := text.lspPosToUtf8Pos r.start;
+let «end» := text.lspPosToUtf8Pos r.«end»;
+let pre := text.source.extract 0 start;
+let post := text.source.extract «end» text.source.bsize;
+(pre ++ newText ++ post).toFileMap
+
 def handleDidChange (p : DidChangeTextDocumentParams) : ServerM Unit := do
 let docId := p.textDocument;
 let changes := p.contentChanges;
@@ -152,7 +158,7 @@ else changes.forM $ fun change =>
   match change with
   | TextDocumentContentChangeEvent.rangeChange (range : Range) (newText : String) => do
     let startOff := oldDoc.text.lspPosToUtf8Pos range.start;
-    let newDocText := oldDoc.text.replaceLspRange range newText;
+    let newDocText := replaceLspRange oldDoc.text range newText;
     (msgLog, newDoc) ← monadLift $
       updateDocument oldDoc startOff newVersion newDocText;
     updateOpenDocuments docId.uri newDoc;