🥮 mooncakes.io

    wget

    wget command for MoonX.

    command-line
    moonx
    wasm
    Download zip
    Author
    Version
    0.2.6
    License
    Apache-2.0
    Repository
    Last updated
    4 days ago
    Downloads
    175

    #wget

    Download HTTP and HTTPS resources for MoonX.

    moonx cli/wget@0.2.6 [OPTION]... [URL]...

    #Supported scope

    Download URLs or URL lists with resume, conditional requests, headers/bodies, retries, redirects, Basic authentication and cookies.

    Unreleased workspace correction: multiple URL failures retain the lowest nonzero transfer exit code, regardless of operand order. A later success keeps that failure status and still writes its response body. HTTP 401 returns 6, including absent credentials and missing or unsupported challenges; other HTTP errors return 8. URL parsing and unsupported input schemes return 1. Network failure 4 takes precedence over authentication failure 6, which takes precedence over HTTP failure 8. The published 0.2.5 coordinate is unchanged; see ADR-0005.

    The workspace recognizes Basic challenges in repeated WWW-Authenticate fields and comma-separated challenge lists, preserving quoted commas and escapes. Final response fields retain their order, casing and repetition after optional-whitespace normalization. Same-name request headers still use the last value; explicit Cookie headers continue across origins.

    URL username/password fields take precedence individually. An omitted URL password can use the configured password; an explicit empty password cannot. HTTP-specific options precede general options regardless of argv order. A missing password prevents Basic authentication; an empty password is valid. Empty URL usernames reject before network access.

    By default URL credentials wait for a challenge independently at each redirect hop. Global option credentials cache challenged hostnames across ports, operands and retries, including failed authentication. URL credentials neither seed nor consume that cache. Explicit Authorization fields remain intact on retry. --auth-no-challenge enables preemptive Basic; attached on/off, yes/no and 1/0 values are case-insensitive, with the last occurrence winning. Binary named-file request bodies replay unchanged. This finite profile does not add Digest/NTLM, netrc, configuration loading, password prompts or arbitrary stdin replay.

    Cookie prefix names retain Wget's ordinary-name behavior. Secure cookies are sent only over HTTPS; curl's loopback exception and prefix/admission/overlay rules do not apply to Wget. Cookie dates follow the finite C-locale profile below; remaining malformed jar grammar stays open under issue #34. The workspace applies the complete pinned ICANN/PRIVATE Public Suffix List to ASCII response cookie domains and hosts, including wildcard and exception rules. Exact public-suffix domains keep Wget's domain scope; imported files retain Wget's separate policy. Published coordinates retain their recorded profile.

    The unpublished --local-encoding=UTF-8 option enables the pinned Unicode 15 IDNA profile for URL hosts, including whole-host transitional retry. Like the fixed oracle build, default raw non-ASCII input retains its spelling; percent-encoded UTF-8 hosts are decoded and converted. Failed IDNA conversion retains the decoded host. ASCII/US-ASCII/ANSI_X3.4-1968 explicitly select the default input profile. Other legacy encodings, invalid UTF-8 bytes, the complete IRI retry dialect and Cookie Domain normalization remain unsupported. Canonical hosts are retained as structured fields across connection setup, redirects and cookie selection. Outgoing Host fields bracket mapped colons and omit the scheme's default port.

    #Options

    OptionDescription
    -q, --quietquiet (no output)
    --local-encoding=ENCSelect UTF-8 or ASCII URL-host input in the unpublished workspace.
    -O, --output-document=FILEwrite documents to FILE
    -o, --output-file=FILElog messages to FILE
    -a, --append-output=FILEappend messages to FILE
    -i, --input-file=FILEdownload URLs found in FILE
    -c, --continueresume after the already written prefix
    -N, --timestampingdo not re-retrieve unchanged files
    -t, --tries=NUMBERset number of retries (0 means unlimited)
    --retry-connrefusedretry even if the connection is refused
    --retry-on-http-error=CODESretry comma-separated HTTP status codes
    --waitretry=SECONDSwait between retries
    --max-redirect=NUMBERmaximum redirects (default 20)
    --content-dispositionhonor Content-Disposition filenames
    --header=STRINGset a request header; last same-name value wins
    --user=USERHTTP Basic user name
    --password=PASSWORDHTTP Basic password
    --http-user, --http-passwordHTTP-specific credentials, before general option fallback
    --auth-no-challenge[=on\|off]control preemptive Basic, including redirects
    --load-cookies=FILEload a Netscape cookie jar
    --header='Cookie: ...'override jar cookies, including on redirects
    --save-cookies=FILEsave persistent cookies
    --keep-session-cookiesalso save session cookies
    --no-cookiesdisable cookie storage
    --method=HTTPMethoduse a supported HTTP method
    --body-data=STRINGsend STRING with an explicit --method
    --body-file=FILEsend FILE with an explicit --method
    --no-check-certificatedisable TLS certificate verification
    --no-proxydo not use proxy environment variables
    -T, --timeout=SECONDSset connect and read timeouts
    --connect-timeout=SECONDSset connection timeout
    --read-timeout=SECONDSset response inactivity timeout
    --idle-timeout=SECONDSextension: response inactivity timeout
    --help / --versionShow help / package version.

    #Limitations

    The workspace checks ASCII response cookie domains using the complete pinned PSL. Admission folds ASCII case and strips leading dots without rewriting the stored key. The fixed Linux Wget build disables NLS: a non-ASCII host or Domain uses its older label heuristic instead of Unicode normalization. Exact matches remain eligible; a parent needs a label boundary and at least two labels. Two-label domains outside .com, .edu, .net, .org, .gov, .mil, .int need more than three UTF-8 bytes in the second-level label. Consequently raw a.公司.cn can accept 公司.cn, while UTF-8 URL conversion changes the actual host and admission result. This is the selected upstream dialect, not browser policy. Wget then discards a matching stored name/domain/port/path even for an unexpired update, after validating that an explicit Path prefixes the request path. It strips one leading Domain dot and retains distinct repeated-dot keys. Request cookie paths use raw prefixes; a missing Path uses the URL directory including its trailing slash. Host-only response cookies from ports other than 80 and 443 retain that port in identity, request selection and jar persistence; explicit Domain cookies and response cookies from those two default ports have no port restriction. An explicit jar port remains restricted, including 80/443. Imported jars use exact uppercase TRUE flags, ignore all comment lines including #HttpOnly_, retain TABs in the last value column, and keep the first domain-chain spelling. Matching name/value pairs are sent once even across different paths. Single-label parent chains are excluded while exact single-label hosts remain eligible; hosts with more than 32 labels do not receive jar cookies. Quoted parameters and ASCII whitespace follow Wget's extractor; Secure ignores an assigned value and HttpOnly is ignored. Empty Expires and Max-Age=-1 reject the field; deletion remains sticky across later expiry attributes. Numeric expiry accepts C-locale decimal/hexadecimal floating prefixes, including incomplete exponents and numeric suffixes. Exactly -1 rejects a Max-Age update; only floating zero requests deletion. A nonzero negative age retains an expired entry and its domain-chain spelling. Expiry equal to the current second remains live; a permanent zero timestamp remains live and can be saved without the session flag. Imported zero timestamps become session cookies. Saved expiry integers use binary64 rounding, matching the fixed Wget profile. Non-finite/out-of-range conversion follows the observed Linux/amd64 oracle explicitly; it is not a portable C guarantee.

    Expires uses Wget's four ordered C-locale date formats (RFC 1123, RFC 850, asctime and Netscape) with English weekday/month names. Two-digit years use the 69 pivot; four-digit year fields accept one to four literal digits. Days 1..31 and seconds 0..61 normalize through UTC calendar arithmetic, even for a nonexistent day within that range. Weekday/date agreement is not checked. A missing timezone, uppercase GMT prefix or signed digit prefix is accepted; timezone offsets are ignored. Other suffixes reject the date, leaving previous expiry operations intact. The timestamp -1 is rejected; zero parses as a valid past date and requests deletion. This follows the fixed Linux/glibc oracle, including ASCII spacing and field-width behavior; it does not claim every libc dialect. Cookie Domain normalization remains separate work.

    • Recursive mirroring, FTP, non-Basic authentication, HSTS and timestamp restoration are unsupported.
    • Content-encoded bodies remain encoded. Consumed stdin request bodies cannot be replayed.
    • Resume retains prior bytes without proving the server's earlier prefix is unchanged. Cookie Domain normalization and complete legacy input encodings remain unsupported.

    #Examples

    moonx cli/wget@0.2.6 -O page.html https://example.com/
moonx cli/wget@0.2.6 -q -O - https://example.com/

    Capability matrix · ADR

    UsageError

    source
    type UsageError

    Options

    source
    type Options

    Source Files