Conversation
|
mmm i'm not sure if the |
|
I am not sure myself. I originally went down a different path, and had another parameter. It then got messy with extra parameter and config. Not terrible, but messy. There is something nice about the URL being self-contained to say "here is what you should take", and it either being a path to an actual file, or to a root that might contain multiple files - could be a directory, or prefix in S3, or even file glob (although globs are not supported right now) - and it indicates what to take. It also makes it a bit easier to implement, but that is a secondary concern. |
|
Mmm ok, but the server will need to have some additional configuration to be able to understand the parameter in the URL and decide what to do. |
What do you mean? Which server will need additional configuration? Not the database, and not the target (smb, s3, file), because those all are parsed internal to mysql-backup anyways. |
|
Mmmm but if it's parsed, why using an url option instead of a command option? I don't understand. |
|
Mainly because
Single URL captures it all. Is there a particular reason to object to it? What is the advantage of any other approach? |
Signed-off-by: Avi Deitcher <avi@deitcher.net>
c193d0c to
f092eb7
Compare
Fixes #436
I am not 100% sure that this will work. The logic is mostly right, but getting tests in place is complex. If you want to try it @ZetaMap and see if it does it, that would be good enough for now.
CI should check that there are no regressions, although restore tests have always been thin.