Problem

According to symbols.map the public core API has 124 symbols¹, which is arguable too much for a key-value API which has as highest goal simplicity.

In particular following areas have many functions and are not simple:

The urgency of this decision is that API can be easily introduced later but we cannot get rid of it after 1.0.

Constraints

Binary data is not a core feature, if needed the plugin system can also work without (properly tagged) binary data in key sets.

The exact API changes are not listed here, because it would just a long list that could just as easily be found in the Git history.

However, the API changes follow these rules:

Remove all functions related to key metadata, except those listed below
Remove all functions related to keyset cursors, as well as ksHead and ksTail
Remove/_Change_ all functions related to (binary) key values as described in (Binary)
Remove all functions that use a user-provided buffer to return keyname/value/etc.
Change/_Add_ keyMeta/keySetMeta to directly read/write the metadata KeySet of a Key
Rename keyGet*Size to key*Size
Add ksRemove function to remove a Key at a specific index
Remove from public API all the helper functions that use elektraMalloc (e.g. elektraFormat), as well as all the helper functions that only enhance standard APIs with additional error checks (e.g. elektraStrCmp)
Remove from public API ksCut, ksDeepDup, ksCopyInternal and other functions that should never have been public
Remove all functions that can be replaced by others (e.g. ksPop)
Rename all functions to start with elektra as described in Elektra Prefix, and use KeySet instead of ks (also applies to names above)

¹ the 124 symbols are (as found by @kodebach):