Skip to content


Subversion checkout URL

You can clone with
Download ZIP
tag: libwww-perl/5.…
Commits on Aug 28, 2001
  1. Release 5.53_96

  2. Release 5.5396

  3. HTTP/1.1 support over SSL.

  4. Call SUPER::configure via _http_socket_configure. This allow subclass…

    …es to override this. Used by the HTTPS support in
  5. Avoid warnings.

Commits on Aug 7, 2001
  1. From: Norton Allen <>

    Subject: [PATCH] Resolve "Short Write"
    Date: Tue, 7 Aug 2001 13:34:24 -0400 (edt)
    The "short write" error reported by is a shorthand for
    "Your operating system has exploited a perfectly legal feature
    that apparently has not been encountered in other operating
    systems to date." That feature is the possibility that syswrite
    might return having only written part of the requested record,
    and I have encountered it under QNX4 for file sizes in excess
    of 10K bytes. This patch does the obvious loop to write out the
      -Norton Allen
  2. From: Blair Zajac <>

    Subject: accept schemantics for HTTP::Daemon and IO::Socket differ
    Date: Mon, 06 Aug 2001 15:02:32 -0700
    In an array context IO::Socket::accept will return the socket and the
    peer returned from ::accept.  HTTP::Daemon::accept always only returns
    the socket, not keeping the same schemantics that you would except by
    reading the HTTP::Daemon manual page.
    Here's a patch to fix this.
  3. Release 5.53_95

  4. Release 5.5395"

  5. Clarifications from "Subject: patch: HTTP::Response minutia" by "Sean…

    … M. Burke" <>
  6. From: "Sean M. Burke" <>

    Subject: fix for language-matching in HTTP::Negotiate
    Date: Sun, 27 May 2001 18:41:49 -0600
    This fixes some odd problems in HTTP::Negotiate's treatment of language
    tags: first, it failed to smash case in one place, and second, it assumed
    that if I accept language A, and you have language AB, then AB is fine.
    It now says that if I accept A, and you have A-B that that's fine.
    Would you believe I meant to make this fix /three/ years ago?
Commits on Aug 6, 2001
  1. Avoid calling epath.

Commits on Aug 2, 2001
  1. Documentation update.

  2. Fix for left over \n at the end of 'message'.

    Also strip of " at %s line %d" where the file name contains spaces.
    Fix by Sean M. Burke <>.
Commits on Jul 21, 2001
  1. If _trivial_http_get got a redirect to non-absolute URI then it

    failed.  Redirects should not really be like that.
Commits on Jul 20, 2001
  1. Fix extract_cookies() behaviour when there is a mix of Set-Cookie and

    Set-Cookie2 headers.  In that case we used to ignore all Set-Cookie
    headers.  Now we only ignore those Set-Cookie headers that reference
    the same cookie as a Set-Cookie2 header.
Commits on May 5, 2001
  1. 5.53_94

  2. Added LWP::Protocol::nogo

  3. From: "Sean M. Burke" <>

    Subject: no LWP::Protocol and LWP::Protocol::nogo
    Date: Sat, 05 May 2001 00:42:32 -0600
    I was grinding my teeth the other day about that fact that if one
    created a CGI that said:
    echo Location: file:///etc/passwd
    and then directed a Perl-based HTTP proxy server to access that
    URL, it would merrily follow redirection to that local file.
    So I created a class, 'LWP::UserAgent::nogo', that one can name
    as the implementor for any protocols (like 'file') that one wishes
    to disable.
  4. From: "Sean M. Burke" <>

    Subject: suggested changes to LWP::UserAgent 1.89
    Date: Sat, 05 May 2001 01:47:58 -0600
    Attached are some suggested changes to LWP::UserAgent -- first
    the whole new file, and then the diffs between
    than and v1.89 -- whichever you find easiest to read.
    They changes several things:
    First, it updates the behavior of redirect_ok according to 2068 (previous
    things said "don't silently redirect POST"; 2068 says "silently redirect
    only HEAD and GET"); then it says not to follow redirect to file:... urls;
    then it allows setting, on a per-UA basis, what request types it'll allow
    redirection for, controlled via the 'redirect_requests' attribute.
    I add per-user attributes for controlling access to protocols; that involved
    changing simple_request to rely on LWP::Protocol::nogo, and also adding
    protocols_allowed and protocols_forbidden accessors.
    That also involved rewriting (and also moving, for sake of the flow of the
    documentation) the is_protocol_supported routine.
    And finally, the thousandth time I started a program with both "use LWP;"
    AND "use HTTP::Request::Common;", I swore I would somehow make it more
    concise for the 95% of requests that are either $ua->request(GET(...))
    or $ua->request(POST(...)), and so I made shortcut methods:
      as a shortcut for
        require HTTP::Request::Common;
    and ditto for head, post, and put.
    I ran the idea past some other programmer folk I know to test whether this
    was just some private mania of mine, but they all liked the idea.  I think
    it also files down one of the pointier corners on the learning curve for
    these modules, and makes the code easier to read: to a neophyte,
      $response = $browser->get($url) is a whole lot clearer than the slightly
      odd-looking $response = $browser->request(GET($url)).
      ("Why do I need to get it, and then request it?", etc.)
Something went wrong with that request. Please try again.