> One of the weirdest things my website does is that it serves everything from a .zip file

whyyyyyyy

why would you do this to yourself

It looks like you're still pushing a bunch of stuff over with git so it's not like you're making it easier to just ftp over one file or something, whyyyyyyy

This is fixable with a small patch, which implements enough seek to work with http.

    func (r *checksumReader) Seek(offset int64, whence int) (int64, error) {
        if whence == io.SeekEnd {
                return int64(r.f.FileHeader.UncompressedSize), nil
        }
        r2, _ := r.f.Open()
        rr := r2.(*checksumReader)
        *r = *rr
        return 0, nil
    }

Another issue is that without seek support, the HTTP server can't implement range requests.

Interesting comment by Russ Cox about this: https://github.com/golang/go/issues/61791#issuecomment-16719...

Reminds me a little of a stored XSS I read about last year.

https://tttang-com.translate.goog/archive/1880/?_x_tr_sl=aut...

Had that same root of not having the mime.types in the container, leading to server-side sniffing of the mime type for the Content-Type header.

It's just a bit interesting the impact such a file can have

Languages do such a bad job at modeling the IO hierarchy ...

One minor point that should be relevant here: there are files that support `rewind` (`seek` at offset `0`) but no other seeking.

This is ironic, as I'm not even sure the authors RSS is setup correctly. My Miniflux instance now has multiple entries for this exact post and in fact the ONLY time I see duplicate posts in there is from the authors site.

abstractions that usually magically work but sometimes don't are the best abstractions (if you're paid by the hour to fix them).

Things that happen when xe stopped to use NixOS