f.lapo.it

Should we provide a 7zip compressed version of the #curl release tarballs?

curl.se/mail/lib-2024-10/0001.…

#cURL
As a Windows user (and somebody who's been using 7-zip since 2002), I really don't see any point in .7z over .tar.xz – the size difference is minimal, and 7-zip handles both types anyway.
It does need to "uncompress it twice" (once to remove .xz, once to extract .tar) though, but that'd be something for 7-zip to solve, not curl.
1
@lapo Yeah, this is something 7-zip should solve (eg. WinRAR already does this).
Wingy mastodon (AP)
@lapo @jernej__s Or just rename the file to a .txz. Then 7-zip should do it in one step.

Unfortunately, it seems not. Neither from the GUI nor from CLI.

% echo a>a
% echo b>b
% tar cJvf test.txz a b
% rm a b
% 7z.exe x test.txz
% ls
test.tar  test.txz
Wingy mastodon (AP)
@lapo @jernej__s Aw that’s sad! I thought that was how 7z worked but it looks like not.
1
maonu mastodon (AP)

@jernej__s 7-zip can offer parallel decompression.

Tar is quite outdated in this regard, in fact!

@maonu @jernej__s I timed my tar xf. It takes 92 milliseconds. For whom is this a problem again?
Matteꙮ Italia mastodon (AP)
agree with this; and if OTOH one wants to reduce file size for someone that only has .zip support, the alternative is to zip twice (no, really); once with "store"/"copy" compression (= no compression), one to actually compress, essentially emulating .tar.gz with .zip (so allowing compression across files); on my machine the generated .zip file (with 7z default options) is 3.9 MB compared to the 6.3 MB of the "regular" .zip.
Questa voce è stata modificata (1 settimana fa)
Jan Johannesson mastodon (AP)
No, not standard. Disk and network are still cheap. Not the guy who's using it.
Marcus Müller mastodon (AP)
don't, honestly. Realistically, the sizes don't matter for human downloaders (also not on the server side, most downloads probably are automated, and will probably be .tar.something), so adding more formats doesn't seem worth the effort. ZIP has one, and honestly, one advantage for Windows users: unlike .tar.xz it can be decompressed with board utilities, but what's the point in that? You end up with a source tarball, and then you install a toolchain that *almost certainly* includes tar.
Mynacol akkoma (AP)
I'm also not convinced at all. Windows 11 now supports all sorts of archive formats out of the box (zip, tar, tar.gz, tar.bz2, tar.xz, tar.zst, 7z, rar). I'd rather look at zstd if you want to add a new (and potentially a better-compressed, fast) format.
@mynacol my zstd tests end up roughly the same size as with xz, so it does not seem worth it.
highvoltage pleroma (AP)
Yeah but zstd decompresses *significantly* faster, which is nice for everyone
my .tar.xz file decompresses in 92 milliseconds for me. I don't believe too many people will get bored before it completes.
Questa voce è stata modificata (1 settimana fa)
René Kijewski mastodon (AP)
manually compressing the .tar file with `xz --extreme --best curl-8.10.1.tar` already reduces the size from 2'766'260 to 2'644'404 bytes. The best 7z compression, `7z a -t7z -m0=lzma -mx=11`, generates roughly the same size: 2'643'431 bytes. So I don't think 7z is needed.
Questa voce è stata modificata (1 settimana fa)
Man2Dev :idle: mastodon (AP)
Considering the email, they do have valid points. The file size difference is significant enough, and the 7z file is more user-friendly in comparison to tar.* for some users. However, it does raise concerns about potential side-channel attacks targeting 7-Zip.
@Man2Dev the diff compared to tar.zx seem insignificant. Why does it raise concerns for attacks?
Man2Dev :idle: mastodon (AP)
Presumably their user base likely prefers user-friendly options like zip. I was comparing the diff between zip and 7z. As for the attack although it migh, not really be practical I was referring to scenarios like the zlib vulnerability.
Questa voce è stata modificata (1 settimana fa)