This seams restrictive to non roman alphabet users? I am not one, but I imagine if I was that I would want to be able to use my default character set to name snapshots. I’m not against the change, just wanted to bring up that drawback.
On Jun 10, 2026 at 9:15:58 PM, Bernardo Botella < [email protected]> wrote: > +1 on this. > > El El lun, 25 may 2026 a las 15:49, Štefan Miklošovič < > [email protected]> escribió: > >> Hello, >> >> We want to validate snapshot names on the server side to prevent >> various issues with it (readers of private ML are aware of them). >> >> It is not enough to cover the most basic validation, we also want to >> validate snapshot length and we want to be sure that snapshots which >> are stored externally (e.g. on AWS S3 or similar) are called in such a >> way that it minimizes the probability of compatibility issues and >> similar. >> >> For AWS, there is a so-called Safe characters set (1) which is >> officially recognized by AWS S3 to be safe to use for the construction >> of object keys. >> >> In (2), I am validating snapshot names based on this character set. >> This safe character set is quite versatile and I already checked with >> other people that restricting snapshot names like this will not impede >> their operations. >> >> Double checking on ML there if people in general are OK with >> validating snapshot names like this? >> >> Regards >> >> (1) >> https://docs.aws.amazon.com/AmazonS3/latest/userguide/object-keys.html#object-key-guidelines >> (2) https://github.com/apache/cassandra/pull/4826/changes >> >
