Replacing the traditional indirect block mapping scheme with an extent mapping scheme, has many benefits, as we have discussed in the previous section. However, changes to the on-disk format that are not backwards compatible are often slow to be adopted by users, for two reasons. First of all, robust e2fsck support sometimes lags the kernel implementation. Secondly, it is generally not possible to mount the filesystem with an older kernel once the filesystem has been converted to use these new features, preventing rollback in case of problems.
Fortunately, there are a number of improvements that can be made to the ext2/3 filesystem without making these sorts of incompatible changes to the on-disk format.
In this section, we will discuss a few of features that are implemented based on the current ext3 filesystem. Section 4.1 describes the effort to reduce the usage of bufferheads structure in ext3; Section 4.2 describes the effort to add delayed allocation without requiring the use of extents; Section 4.3 discusses the work to add multiple block allocation; Section 4.4 describes asynchronous file unlink and truncate; Section 4.5 describes a feature to allow more than 32000 subdirectories; and Section 4.6 describes a feature to allow multiple threads to concurrently create/rename/link/unlink files in a single directory.