|
MSDOS had it right with 8.3 in all caps
|
# ? Feb 1, 2024 11:13 |
|
|
# ? May 24, 2024 07:21 |
|
it is not like it is some intractable problem to do unicode-defined normalization, case folding and e.g. collation either. it is hardly a necessary feature, but it's nice and not *that* complex.
|
# ? Feb 1, 2024 11:26 |
|
it’s not hard, but it is also locale-dependent, which is inherently at odds with being something like a filesystem which is meant to reflect the same data for different users
|
# ? Feb 1, 2024 13:07 |
|
I don’t think it would be that unreasonable to say “this is the locale that /whatever is mounted as”, if you wanted the correct case folding in exchange I wonder what the performance of that would be like. maybe we’d get RISC-V instruction extensions for doing the collation and folded comparison, requiring a gig of fast cache just for the tables
|
# ? Feb 1, 2024 13:45 |
|
for collation locale is kind of a big deal, but the filesystem does not usually handle that part (afaik at least?). for normalization/case folding i think you indeed do just fix a sensible default on normalization/case-folding. debate is then mostly whether "sensible" is just 1:1 (i.e. just memcmp blindly), fixing a unicode table (possibly tweaked), and if so if you include various casing-"style" normalization. windows does do the latter. i can't say i care a ton, but i think it overall matters more for some languages that others. e.g. greek has the pleasure of having not only U+03c1 (Small Greek Rho) and U+03a1 (Capital Greek Rho), but *also* U+03f1 (Greek Rho Symbol), and that gets a touch on the side of bullshit to have people interface directly with.
|
# ? Feb 1, 2024 14:03 |
|
goon project: invent a language where nul byte is meaningful
|
# ? Feb 1, 2024 14:14 |
|
xargs -0
|
# ? Feb 1, 2024 14:23 |
|
Sweeper posted:goon project: invent a language where nul byte is meaningful *the zen master kicks you in the balls*
|
# ? Feb 1, 2024 14:41 |
|
I've seen apps just url encode file names instead of thinking about encodings and I can't say they're wrong
|
# ? Feb 1, 2024 14:44 |
Filesystems/OSs should just give every file a UUID and then use that. Then, allow users to set the display names to whatever they want. Users wanna have 50 different files on their desktop, all called "temp.txt" or whatever? I say, let them!
|
|
# ? Feb 1, 2024 18:27 |
|
export LC_ALL=C
|
# ? Feb 1, 2024 18:28 |
|
VikingofRock posted:Filesystems/OSs should just give every file a UUID and then use that. Then, allow users to set the display names to whatever they want. Users wanna have 50 different files on their desktop, all called "temp.txt" or whatever? I say, let them! this would make revision control systems, backups, and even just interacting with any sort of archive format rather exciting
|
# ? Feb 1, 2024 18:44 |
|
let a thousand file-overwrite vulnerabilities bloom
|
# ? Feb 1, 2024 18:52 |
Internet Janitor posted:this would make revision control systems, backups, and even just interacting with any sort of archive format rather exciting Eh, we could all use a little more excitement in our lives
|
|
# ? Feb 1, 2024 18:52 |
|
Vanadium posted:I've seen apps just url encode file names instead of thinking about encodings and I can't say they're wrong I can, because URL encoding is defined on characters, and file names can contain sequences that aren’t valid characters
|
# ? Feb 1, 2024 19:13 |
|
all files should be named as the memory address they point to. Users shouldn’t be given an option to name them. Good ol 0x549badc60381c2361d53ce2195e926f7. And yes, the file names should change if the memory address changes.
|
# ? Feb 1, 2024 19:20 |
|
files should be named with the timestamp at which they’re permitted to be opened. before that moment, permission denied. after that moment, immediately deleted strict monotonic clock, of course
|
# ? Feb 1, 2024 19:33 |
|
a filesystem that incorporated a concept of (configurable) expiration dates would be kind of interesting, and in some cases even useful
|
# ? Feb 1, 2024 19:41 |
|
it's kinda like a snapchat operating system
|
# ? Feb 1, 2024 19:54 |
|
if only posts could do the same
|
# ? Feb 1, 2024 20:36 |
|
echinopsis posted:why not have the real file name and file display name be different things we can even store the file display name in the resource fork associated with the file!
|
# ? Feb 1, 2024 20:52 |
|
operating systems should provide virtual disk services in the same way that they provide virtual memory services. carve the disk up into 64MB (or whatever size) pages and allocate logical volumes to per-user user-space file system implementations to do with as they please.
|
# ? Feb 1, 2024 21:15 |
|
i mean, they already do that linux lvm will happily export device nodes carving up a disk into virtual disks and there's nothing stopping userspace from opening one of those block devices and going ham much finer granularity than 64MB too
|
# ? Feb 1, 2024 21:51 |
|
would your filesystem and os need to keep up to date with new unicode versions?
|
# ? Feb 1, 2024 23:29 |
|
VikingofRock posted:Filesystems/OSs should just give every file a UUID and then use that. Then, allow users to set the display names to whatever they want. Users wanna have 50 different files on their desktop, all called "temp.txt" or whatever? I say, let them! hell, every file is 24 digit alphabuermic random number w
|
# ? Feb 1, 2024 23:37 |
|
echinopsis posted:hell, every file is 24 digit alphabuermic random number w
|
# ? Feb 1, 2024 23:38 |
|
redleader posted:would your filesystem and os need to keep up to date with new unicode versions? Windows embeds the case-folding table into NTFS at filesystem creation, and then doesn't use it at runtime. iirc macOS has changed it's normalization algorithm incompatibly as Unicode has changed incompatibly (or maybe it was the reverse, Unicode changed incompatibly and macOS kept the obsolete Unicode normalization algorithm) in conclusion, normalizing filenames is stupid
|
# ? Feb 1, 2024 23:46 |
|
redleader posted:im leaving that in the well-qualified hands of the unicode committee yes, the people who brought us han unification surely won't gently caress this up in any way
|
# ? Feb 1, 2024 23:56 |
|
Visions of Valerie posted:yes, the people who brought us han unification surely won't gently caress this up in any way wasn’t there a Han unification effort before Unicode?
|
# ? Feb 2, 2024 00:21 |
|
pseudorandom name posted:Windows embeds the case-folding table into NTFS at filesystem creation, and then doesn't use it at runtime. what? why?
|
# ? Feb 2, 2024 00:29 |
|
Subjunctive posted:wasn’t there a Han unification effort before Unicode?
|
# ? Feb 2, 2024 00:32 |
|
Should have done Latin-Greek-Cyrillic unification as well while they were at it D, Δ, Д are all basically the same letter, if the user really about that character's exact appearance then they can choose a region-appropriate font
|
# ? Feb 2, 2024 04:27 |
|
Д is just A
|
# ? Feb 2, 2024 04:46 |
|
but then how will i make "put it in н" jokes
|
# ? Feb 2, 2024 04:48 |
|
Visions of Valerie posted:yes, the people who brought us han unification surely won't gently caress this up in any way actually han unification is fine, according to white people on the internet Sapozhnik posted:Should have done Latin-Greek-Cyrillic unification as well while they were at it ah, no, this is different from han unification because
|
# ? Feb 2, 2024 05:36 |
|
han unification in unicode was designed and directed by CJK-JRG, which is staffed with experts from China, Japan, and Korea. I assume that they at least aren’t all white
|
# ? Feb 2, 2024 05:38 |
|
this gives me a headache
|
# ? Feb 2, 2024 06:10 |
|
VikingofRock posted:Filesystems/OSs should just give every file a UUID and then use that
|
# ? Feb 2, 2024 06:15 |
|
redleader posted:ah, no, this is different from han unification because e.g. KOI8-R contains both A and Д and you need to be able to losslessly roundtrip from KOI8-R to Unicode and then back to KOI8-R
|
# ? Feb 2, 2024 06:52 |
|
|
# ? May 24, 2024 07:21 |
|
its the same reason Unicode contains a bunch of precomposed characters even though you can make them using combining sequences and also why unicode contains both precomposed and combining forms of hangul
|
# ? Feb 2, 2024 07:00 |