diff --git a/DESCRIPTION b/DESCRIPTION
index 5e1cd3751..4aa682d9f 100644
--- a/DESCRIPTION
+++ b/DESCRIPTION
@@ -16,8 +16,9 @@ Description: Pretty-prints R code without changing the user's formatting
 License: MIT + file LICENSE
 URL: https://github.com/r-lib/styler, https://styler.r-lib.org
 BugReports: https://github.com/r-lib/styler/issues
+Depends:
+    R (>= 3.4.0)
 Imports: 
-    backports (>= 1.1.0),
     cli (>= 1.1.0),
     glue,
     magrittr (>= 2.0.0),
@@ -29,7 +30,6 @@ Imports:
     tibble (>= 1.4.2),
     tools,
     withr (>= 1.0.0),
-    xfun (>= 0.1)
 Suggests: 
     data.tree (>= 0.1.6),
     digest,
diff --git a/NEWS.md b/NEWS.md
index ea1b17161..ef6ebddb9 100644
--- a/NEWS.md
+++ b/NEWS.md
@@ -9,6 +9,9 @@
   (also used by {precommit}).
 * rename default branch to main (#859).
 * Fix argument name `filetype` in Example for `style_dir()` (#855).
+* Bump minimal R requirement to 3.4 in line with the [tidyverse](https://www.tidyverse.org/blog/2019/04/r-version-support/), which 
+  allowed to remove the dependency at {backports} and some exception handling.
+* Remove dependency on {xfun} (#866).
 
 # styler 1.6.2
 
diff --git a/R/io.R b/R/io.R
index 9caad1a67..0c588500e 100644
--- a/R/io.R
+++ b/R/io.R
@@ -42,7 +42,7 @@ transform_utf8_one <- function(path, fun, dry) {
         } else if (dry == "on") {
           # don't do anything
         } else if (dry == "off") {
-          xfun::write_utf8(new, path)
+          write_utf8(new, path)
         } else {
           # not implemented
         }
@@ -114,3 +114,10 @@ read_utf8_bare <- function(con, warn = TRUE) {
 invalid_utf8 <- function(x) {
   which(!is.na(x) & is.na(iconv(x, "UTF-8", "UTF-8")))
 }
+
+#' Drop-in replacement for `xfun::write_utf8()`
+#' @keywords internal
+write_utf8 <- function(text, con, ...) {
+  withr::local_options(encoding = "native.enc")
+  writeLines(enc2utf8(text), con, ..., useBytes = TRUE)
+}
diff --git a/R/parse.R b/R/parse.R
index 13a4788e6..a82c60637 100644
--- a/R/parse.R
+++ b/R/parse.R
@@ -127,7 +127,6 @@ add_id_and_short <- function(pd) {
 #' @importFrom magrittr or
 #' @keywords internal
 ensure_correct_txt <- function(pd, text) {
-  ensure_valid_pd(pd)
   is_problematic_text <- or(
     is_insufficiently_parsed_string(pd),
     is_insufficiently_parsed_number(pd)
@@ -173,30 +172,6 @@ ensure_correct_txt <- function(pd, text) {
     arrange_pos_id()
 }
 
-#' Ensure that the parse data is valid
-#'
-#' Test whether all non-terminals have at least one child and throw an error
-#' otherwise. As this is check is rather expensive, it is only
-#' carried out for configurations we have good reasons to expect problems.
-#' @param pd A parse table.
-#' @importFrom rlang abort
-#' @keywords internal
-ensure_valid_pd <- function(pd) {
-  if (getRversion() < "3.2") {
-    non_terminals <- pd %>%
-      filter(terminal == FALSE)
-    valid_pd <- non_terminals$id %>%
-      map_lgl(~ .x %in% pd$parent) %>%
-      all()
-    if (!valid_pd) {
-      abort(paste(
-        "The parse data is not valid and the problem is most likely related",
-        "to the parser in base R. Please install R >= 3.2 and try again."
-      ))
-    }
-  }
-  TRUE
-}
 
 #' Identify strings that were not fully parsed
 #'
diff --git a/R/roxygen-examples-find.R b/R/roxygen-examples-find.R
index 6b0be661e..941414b14 100644
--- a/R/roxygen-examples-find.R
+++ b/R/roxygen-examples-find.R
@@ -22,7 +22,7 @@ identify_start_to_stop_of_roxygen_examples_from_text <- function(text) {
 }
 
 identify_start_to_stop_of_roxygen_examples <- function(path) {
-  content <- xfun::read_utf8(path)
+  content <- read_utf8_bare(path)
   identify_start_to_stop_of_roxygen_examples_from_text(content)
 }
 
diff --git a/R/testing.R b/R/testing.R
index 9754b9a16..1238b8e6f 100644
--- a/R/testing.R
+++ b/R/testing.R
@@ -110,7 +110,7 @@ transform_and_check <- function(in_item, out_item,
                                 write_tree = NA,
                                 out_tree = "_tree", ...) {
   write_tree <- set_arg_write_tree(write_tree)
-  read_in <- xfun::read_utf8(in_item)
+  read_in <- read_utf8_bare(in_item)
   if (write_tree) {
     create_tree(read_in) %>%
       write.table(out_tree,
diff --git a/R/transform-files.R b/R/transform-files.R
index b8cf90f0c..619b134e5 100644
--- a/R/transform-files.R
+++ b/R/transform-files.R
@@ -232,7 +232,7 @@ parse_transform_serialize_r <- function(text,
     return("")
   }
   transformers <- transformers_drop(
-    if (getRversion() < 3.4) text else pd_nested$text[!pd_nested$is_cached],
+    pd_nested$text[!pd_nested$is_cached],
     transformers
   )
 
diff --git a/R/ui-styling.R b/R/ui-styling.R
index 9117fa3f0..bd035c336 100644
--- a/R/ui-styling.R
+++ b/R/ui-styling.R
@@ -312,7 +312,7 @@ prettify_any <- function(transformers,
 #' @inheritSection style_pkg Round trip validation
 #' @examples
 #' file <- tempfile("styler", fileext = ".R")
-#' xfun::write_utf8("1++1", file)
+#' writeLines("1++1", file)
 #'
 #' # the following is identical (because of ... and defaults),
 #' # but the first is most convenient:
@@ -325,7 +325,7 @@ prettify_any <- function(transformers,
 #' # name levels explicitly to not style less invasive levels
 #' style_file(file, scope = I(c("tokens", "spaces")), strict = TRUE)
 #'
-#' xfun::read_utf8(file)
+#' readLines(file)
 #' unlink(file)
 #' @family stylers
 #' @export
diff --git a/R/zzz.R b/R/zzz.R
index 427fc0083..1464cd2dd 100644
--- a/R/zzz.R
+++ b/R/zzz.R
@@ -1,5 +1,4 @@
 .onLoad <- function(libname, pkgname) {
-  backports::import(pkgname, "trimws")
   op <- options()
   op.styler <- list(
     styler.addins_style_transformer = "styler::tidyverse_style()",
@@ -37,8 +36,8 @@ ask_to_switch_to_non_default_cache_root <- function(ask = interactive()) {
       "options(styler.cache_root = \"styler\")\n\nin your `.Rprofile`. This ",
       "message will only be displayed once in a while.\n"
     ))
+    options(styler.cache_root = "styler")
   }
-  options(styler.cache_root = "styler")
 }
 
 remove_old_cache_files <- function() {
diff --git a/man/ensure_valid_pd.Rd b/man/ensure_valid_pd.Rd
deleted file mode 100644
index c6e1cd8d9..000000000
--- a/man/ensure_valid_pd.Rd
+++ /dev/null
@@ -1,17 +0,0 @@
-% Generated by roxygen2: do not edit by hand
-% Please edit documentation in R/parse.R
-\name{ensure_valid_pd}
-\alias{ensure_valid_pd}
-\title{Ensure that the parse data is valid}
-\usage{
-ensure_valid_pd(pd)
-}
-\arguments{
-\item{pd}{A parse table.}
-}
-\description{
-Test whether all non-terminals have at least one child and throw an error
-otherwise. As this is check is rather expensive, it is only
-carried out for configurations we have good reasons to expect problems.
-}
-\keyword{internal}
diff --git a/man/style_file.Rd b/man/style_file.Rd
index 3c58e917f..95fec034e 100644
--- a/man/style_file.Rd
+++ b/man/style_file.Rd
@@ -97,7 +97,7 @@ See section 'Warning' for a good strategy to apply styling safely.
 
 \examples{
 file <- tempfile("styler", fileext = ".R")
-xfun::write_utf8("1++1", file)
+writeLines("1++1", file)
 
 # the following is identical (because of ... and defaults),
 # but the first is most convenient:
@@ -110,7 +110,7 @@ style_file(file, scope = "indention", strict = TRUE)
 # name levels explicitly to not style less invasive levels
 style_file(file, scope = I(c("tokens", "spaces")), strict = TRUE)
 
-xfun::read_utf8(file)
+readLines(file)
 unlink(file)
 }
 \seealso{
diff --git a/man/write_utf8.Rd b/man/write_utf8.Rd
new file mode 100644
index 000000000..9955e86e2
--- /dev/null
+++ b/man/write_utf8.Rd
@@ -0,0 +1,12 @@
+% Generated by roxygen2: do not edit by hand
+% Please edit documentation in R/io.R
+\name{write_utf8}
+\alias{write_utf8}
+\title{Drop-in replacement for \code{xfun::write_utf8()}}
+\usage{
+write_utf8(text, con, ...)
+}
+\description{
+Drop-in replacement for \code{xfun::write_utf8()}
+}
+\keyword{internal}
diff --git a/tests/testthat/test-parsing.R b/tests/testthat/test-parsing.R
index 08d3d410b..73d0db1cf 100644
--- a/tests/testthat/test-parsing.R
+++ b/tests/testthat/test-parsing.R
@@ -21,15 +21,13 @@ test_that("repreated parsing solves wrong parent assignment", {
     "R -q -e \"styler::cache_deactivate(); styler::style_file(\\\"", path_temp, "\\\")\""
   )
   calls_sys(sys_call, intern = FALSE, ignore.stdout = TRUE, ignore.stderr = TRUE)
-  ref <- xfun::read_utf8(testthat_file("parsing", "repeated_parsing-out.R"))
-  result <- xfun::read_utf8(path_temp)
+  ref <- read_utf8_bare(testthat_file("parsing", "repeated_parsing-out.R"))
+  result <- read_utf8_bare(path_temp)
   expect_equal(ref, result)
   unlink(dir)
 })
 
 test_that("long strings are parsed correctly", {
-  if (getRversion() < "3.2") skip("skip on R < 3.2 because of parsing problems")
-
   expect_warning(
     test_collection("parsing", "long_strings", transformer = style_text),
     NA
@@ -47,20 +45,7 @@ test_that("0x number representation is preserved with(out) L", {
 })
 
 
-test_that("issues with parsing long strings on R 3.1 can be detected", {
-  if (getRversion() >= "3.2") {
-    skip("skip on R >= 3.2 because parsing probmes don't appear")
-  }
-  expect_error(
-    test_collection("parsing", "long_strings", transformer = style_text),
-    "install R .* 3.2"
-  )
-})
-
-
 test_that("CRLF EOLs fail with informative error", {
-  skip_if(getRversion() < "3.4")
-
   expect_error(
     style_text("glück <- 3\r\n glück + 1"),
     "Please change the EOL character in your editor to Unix style and try again."