BreakingChanges: announce switch to "reftable" format

The "reftable" format has come a long way and has matured nicely since
it has been merged into git via 57db2a0 (refs: introduce reftable
backend, 2024-02-07). It fixes longstanding issues that cannot be fixed
with the "files" format in a backwards-compatible way and performs
significantly better in many use cases.

Announce that we will switch to the "reftable" format in Git 3.0 for
newly created repositories and wire up the change, hidden behind the
WITH_BREAKING_CHANGES preprocessor define.

This switch is dependent on support in the larger Git ecosystem. Most
importantly, libraries like JGit, libgit2 and Gitoxide should support
the reftable backend so that we don't break all applications and tools
built on top of those libraries.

Signed-off-by: Patrick Steinhardt <[email protected]>
Signed-off-by: Junio C Hamano <[email protected]>

Author: pks-t

Documentation/BreakingChanges.adoc

@@ -118,6 +118,53 @@ Cf. <[email protected]>,
 <[email protected]>,
 <CA+EOSBncr=4a4d8n9xS4FNehyebpmX8JiUwCsXD47EQDE+DiUQ@mail.gmail.com>.
 
+* The default storage format for references in newly created repositories will
+  be changed from "files" to "reftable". The "reftable" format provides
+  multiple advantages over the "files" format:
++
+  ** It is impossible to store two references that only differ in casing on
+     case-insensitive filesystems with the "files" format. This issue is common
+     on Windows and macOS platforms. As the "reftable" backend does not use
+     filesystem paths to encode reference names this problem goes away.
+  ** Similarly, macOS normalizes path names that contain unicode characters,
+     which has the consequence that you cannot store two names with unicode
+     characters that are encoded differently with the "files" backend. Again,
+     this is not an issue with the "reftable" backend.
+  ** Deleting references with the "files" backend requires Git to rewrite the
+     complete "packed-refs" file. In large repositories with many references
+     this file can easily be dozens of megabytes in size, in extreme cases it
+     may be gigabytes. The "reftable" backend uses tombstone markers for
+     deleted references and thus does not have to rewrite all of its data.
+  ** Repository housekeeping with the "files" backend typically performs
+     all-into-one repacks of references. This can be quite expensive, and
+     consequently housekeeping is a tradeoff between the number of loose
+     references that accumulate and slow down operations that read references,
+     and compressing those loose references into the "packed-refs" file. The
+     "reftable" backend uses geometric compaction after every write, which
+     amortizes costs and ensures that the backend is always in a
+     well-maintained state.
+  ** Operations that write multiple references at once are not atomic with the
+     "files" backend. Consequently, Git may see in-between states when it reads
+     references while a reference transaction is in the process of being
+     committed to disk.
+  ** Writing many references at once is slow with the "files" backend because
+     every reference is created as a separate file. The "reftable" backend
+     significantly outperforms the "files" backend by multiple orders of
+     magnitude.
+  ** The reftable backend uses a binary format with prefix compression for
+     reference names. As a result, the format uses less space compared to the
+     "packed-refs" file.
++
+Users that get immediate benefit from the "reftable" backend could continue to
+opt-in to the "reftable" format manually by setting the "init.defaultRefFormat"
+config. But defaults matter, and we think that overall users will have a better
+experience with less platform-specific quirks when they use the new backend by
+default.
++
+A prerequisite for this change is that the ecosystem is ready to support the
+"reftable" format. Most importantly, alternative implementations of Git like
+JGit, libgit2 and Gitoxide need to support it.
+
 === Removals
 
 * Support for grafting commits has long been superseded by git-replace(1).

help.c

@@ -810,6 +810,8 @@ void get_version_info(struct strbuf *buf, int show_build_options)
 			    SHA1_UNSAFE_BACKEND);
 #endif
 		strbuf_addf(buf, "SHA-256: %s\n", SHA256_BACKEND);
+		strbuf_addf(buf, "default-ref-format: %s\n",
+			    ref_storage_format_to_name(REF_STORAGE_FORMAT_DEFAULT));
 	}
 }
 

repository.h

@@ -20,6 +20,12 @@ enum ref_storage_format {
 	REF_STORAGE_FORMAT_REFTABLE,
 };
 
+#ifdef WITH_BREAKING_CHANGES /* Git 3.0 */
+# define REF_STORAGE_FORMAT_DEFAULT REF_STORAGE_FORMAT_REFTABLE
+#else
+# define REF_STORAGE_FORMAT_DEFAULT REF_STORAGE_FORMAT_FILES
+#endif
+
 struct repo_path_cache {
 	char *squash_msg;
 	char *merge_msg;

setup.c

@@ -2541,6 +2541,8 @@ static void repository_format_configure(struct repository_format *repo_fmt,
 			repo_fmt->ref_storage_format = ref_format;
 	} else if (cfg.ref_format != REF_STORAGE_FORMAT_UNKNOWN) {
 		repo_fmt->ref_storage_format = cfg.ref_format;
+	} else {
+		repo_fmt->ref_storage_format = REF_STORAGE_FORMAT_DEFAULT;
 	}
 	repo_set_ref_storage_format(the_repository, repo_fmt->ref_storage_format);
 }

t/t0001-init.sh

@@ -658,6 +658,17 @@ test_expect_success 'init warns about invalid init.defaultRefFormat' '
 	test_cmp expected actual
 '
 
+test_expect_success 'default ref format' '
+	test_when_finished "rm -rf refformat" &&
+	(
+		sane_unset GIT_DEFAULT_REF_FORMAT &&
+		git init refformat
+	) &&
+	git version --build-options | sed -ne "s/^default-ref-format: //p" >expect &&
+	git -C refformat rev-parse --show-ref-format >actual &&
+	test_cmp expect actual
+'
+
 backends="files reftable"
 for format in $backends
 do