
On 11.08.2016 19:43, Cole Robinson wrote:
On 08/11/2016 12:02 PM, Michal Privoznik wrote:
Since ages filesystems allowed to have space characters in filenames and even directory names. In fact, on all major filesystems out there you can have whatever character you like except NULL. There's no reason why we should forbid users to not have spaces in their filenames. Moreover, if we do that only on RNG schema level while our XML parser/formatter crunches that happily.
There's a fedora bug about this particular issue:
Ah, thank you for pointing that out. I often forget about those.
But for example this range still rejects other valid characters as well, like unicode รก . So maybe rather than a whitelist, we go the opposite way and make this a minimal blacklist, or drop the validation entirely. Unless there's some designated way to handle regex validation for unicode...
Yeah, I guess we can drop the validation completely. I see couple of reasons to do that: a) all modern filesystems allow users to have whatever character they want in the file name (except NULL) [1] b) we must not think about disk sources as UNIX style paths. I mean, there are some hypervisors (like ESX) where disk source is just a name that is translated by hypervisor then into a path that it understands. c) it's okay if we have wider schema but narrower parser. I mean, if schema allows something that is later reject by parser. Will post v2. Thank you. Michal 1: https://en.wikipedia.org/wiki/Comparison_of_file_systems#Limits