This is a similar is #190 but for regular rather than Xapian indexing. This is due to the word field in eprint__rindex being a VARCHAR(128). That is not unreasonable as it is unlikely that a word that might be searched on would be longer than that number of characters. The issue is not due to multi-bytes words, it is due to MariaDB 10.3+ complaining about inserting words longer than 128 characters to eprint__rindex rather than just truncating. Neither truncating or error is appropriate. It is unlikely to be useful to index a word greater than 128 characters even it if is possible, so these should be excluded from indexing.