ZERO_LENGTH RANGE status code

[carpawell]

We have OUT_OF_RANGE status code but if a request is with a zero-length, we return a general error cause that is not OUT_OF_RANGE in fact (but on our CLI, SDK, and SN levels it is prohibited).

Do we need some ZERO_LENGTH response code? Note: the current API does not say it MUST be prohibited.

[roman-khimov]

OUT_OF_RANGE doesn't seem to be far off for this.

[carpawell]

So you mean using OUT_OF_RANGE for zero-length? We do not do it now.

[roman-khimov]

Yeah, it should be good enough. @cthulhu-rider?

[cthulhu-rider]

it's documented like this, so, according to them, we may respond with this status on any invalid range

at the same time, for example, gRPC says that Unlike INVALID_ARGUMENT...
(imagine #269) this error indicates a problem that may be fixed if the system state changes

in total, i vote for BAD_REQUEST cuz it more like it's nonsense in all cases while OUT_OF_RANGE is more about object exists (!) but u are out of its range (technically we may respond to [0,0] range with empty data)

P.S. zero RANGE request is the only way to check object existence w/o any object data transmission, feature huh? i don't think so, use HEAD instead

[carpawell]

Agree that OUT_OF_RANGE is more like an incorrect range (does not fit an object) while [X:0] is about a senseless range.