Changelog

Time flies!!

3.0.1 (2026-11-01)

  • Default Accept-Encoding is gzip, deflate. The previous value included *, so a server could send an encoding the client cannot decode.

  • Header names are matched case-insensitively. An existing name keeps its spelling and the new value replaces it.

3.0.0 (2026-08-24)

Backwards-incompatible:

  • Session.putcookie() and init cookies are host-scoped. A cookie without domain is bound to the host of the next request and is not sent to other hosts.

  • Response.text decodes with Content-Type charset or UTF-8. It no longer guesses gbk/big5 via mb_code().

  • Dropped Python 2 shims: basestring, b, u, and Response.next.

2.4.1 (2026-08-24)

  • no_proxy matches IPv6 addresses and CIDR, *, and domain suffixes without treating notexample.com as example.com

  • IPv6 Host headers use RFC 3986 brackets ([::1]:8080)

2.4.0 (2026-08-24)

  • Response.encoding and Response.iter_lines()

  • files values may be (filename, content, content_type)

2.3.0 (2026-08-24)

  • Response.text uses the charset from Content-Type when present

  • Multipart field names and filenames are quoted; file parts use mimetypes for Content-Type

2.2.1 (2026-08-24)

  • Classifiers for Python 3.14 and 3.15

2.2.0 (2026-08-24)

  • HTTPS through an HTTP proxy uses CONNECT (HTTPSConnection.set_tunnel)

  • Optional ssl_context for HTTPS

  • Response.ok and Response.raise_for_status()

2.1.0 (2026-08-24)

  • Session cookies from responses are stored in a cookiejar and only sent back to matching hosts. Cookies set with Session.putcookie() or the cookies argument are still sent to every host.

  • Add json= to request() / post() / fetch()

  • Replace the 6000-line user-agent list with a small set of current browsers

2.0.3 (2026-08-24)

  • Require Python >= 3.8 (CI can no longer install 3.5–3.7)

  • Follow 308 redirects; 307/308 keep the original method and body

  • 301/302/303 redirects no longer resend the request body

  • Strip Authorization and Cookie on cross-origin redirects

  • Fix Proxy-Authorization password on redirects

  • Honor NO_PROXY when following redirects

  • Close HTTP connections on the response and on errors

  • Parse multiple Set-Cookie headers

  • Export Timeout; reject mutable default arguments

  • Replace the Sina network compression tests with local fixtures

2.0.2 (2026-08-24)

  • Publish to PyPI via GitHub Actions Trusted Publishing

  • Automatically create GitHub Releases with changelog and generated notes

2.0.1 (2023-08-04)

  • Fixed: formdata fieldname

  • Move import zlib to top level

  • Change urlparse to urlsplit

  • setup.py python_required=”>=3.5”

2.0.0 (2023-07-29)

  • Drop support for Python 2.6, 2.7, 3.2, 3.3 and 3.4. urlfetch requires Python >= 3.5 from now on.

  • Unless specified, assume it’s http proxy

  • Migrate CI to github actions

1.2.2 (2020-07-11)

  • Fixed: proxy scheme bug

1.2.1 (2020-04-29)

  • Fixed bug: passing context to HTTPSConnection

1.2.0 (2020-04-29)

  • (Contributed by @chmoder) Added validate_certificate to skip validating certificates.

1.1.3 (2019-10-11)

  • (Contributed by @chmoder) Define HTTP request methods as string constants, now we can use urlfetch.request(urlfetch.POST, …).

1.1.2 (2019-03-27)

Small optimizations:

  • Larger chunk when reading response

  • Read chuncks into list and then join them to bytes

  • Close response when exception occurs

1.1.1 (2018-12-20)

  • Updated user-agent list.

1.1.0 (2018-11-16)

New features:

  • Support source_address

  • Support no_proxy environment variable

1.0.3 (2018-01-03)

Improvements:

  • Run tests against Python 3.5 3.6 3.7 and PyPy.

  • Try to deal with data_files paths.

  • Some minor changes regarding coding style.

1.0.2 (2015-04-29)

Fixes:

  • python setup.py test causes SandboxViolation.

Improvements:

  • python setup.py test handles dependencies automatically.

  • random_useragent(): check if urlfetch.useragents.list exists at the import time.

1.0.1 (2015-01-31)

Fixes:

  • urlfetch.Response.history of a redirected response and its precedent responses should be different.

Improvements:

  • Simplified some code.

  • Added some tests.

1.0 (2014-03-22)

New features:

Backwards-incompatible changes:

  • Remove raw_header and raw_response.

  • random_useragent() now takes a single filename as parameter. It used to be a list of filenames.

  • No more .title() on request headers’ keys.

  • Exceptions are re-designed. socket.timeout now is Timeout, …, see section Exceptions in Reference for more details.

Fixes:

  • Parsing links: If Link header is empty, [] should be returned, not [{'url': ''}].

  • Http request’s Host header should include the port. Using netloc as the http host header is wrong, it could include user:pass.

  • Redirects: Host in reqheaders should be host:port.

  • Streaming decompress not working.

0.6.2 (2014-03-22)

Fix:

  • Http request’s host header should include the port. Using netloc as the http host header is wrong, it could include user:pass.

0.6.1 (2014-03-15)

Fix:

  • Parsing links: If Link header is empty, [] should be returned, not [{'url': ''}].

0.6 (2013-08-26)

Change:

  • Remove lazy response introduced in 0.5.6

  • Remove the dump, dumps, load and loads methods of urlfetch.Response

0.5.7 (2013-07-08)

Fix:

  • Host header field should include host and port

0.5.6 (2013-07-04)

Feature:

  • Lay response. Read response when you need it.

0.5.5 (2013-06-07)

Fix:

0.5.4.2 (2013-03-31)

Feature:

  • urlfetch.Response.link, links parsed from HTTP Link header.

Fix:

  • Scheme doesn’t correspond to the new location when following redirects.

0.5.4.1 (2013-03-05)

Fix:

0.5.4 (2013-02-28)

Feature:

  • HTTP Proxy-Authorization.

Fix:

0.5.3.1 (2013-02-01)

Fix:

0.5.3 (2013-02-01)

Feature:

Fix:

  • urlfetch.mb_code() may silently return incorrect result, since the encode errors are replaced, it should be decode properly and then encode without replace.

0.5.2 (2012-12-24)

Feature:

  • random_useragent() can accept list/tuple/set params and can accept more than one params which specify the paths to check and read from. Below are some examples:

    >>> ua = random_useragent('file1')
>>> ua = random_useragent('file1', 'file2')
>>> ua = random_useragent(['file1', 'file2'])
>>> ua = random_useragent(['file1', 'file2'], 'file3')

Fix:

0.5.1 (2012-12-05)

Fix:

  • In some platforms urlfetch.useragents.list located in wrong place.

  • random_useragent() will never return the first line.

  • Typo in the description of urlfetch.useragents.list (the first line).

0.5.0 (2012-08-23)

  • Redirects support. Parameter max_redirects specify the max redirects allowed within a request. Default is 0, which means redirects are not allowed.

  • Code cleanups

0.4.3 (2012-08-17)

  • Add params parameter, params is dict or string to attach to request url as querysting.

  • Gzip and deflate support.

0.4.2 (2012-07-31)

  • HTTP(S) proxies support.

0.4.1 (2012-07-04)

  • Streaming support.

0.4.0 (2012-07-01)

0.3.6 (2012-06-08)

  • Simplify code

  • Trace method without data and files, according to RFC2612

  • urlencode(data, 1) so that urlencode({'param': [1,2,3]}) => 'param=1&param=2&param=3'

0.3.5 (2012-04-24)

  • Support specifying an IP for the request host, useful for testing API.

0.3.0 (2012-02-28)

  • Python 3 compatible

0.2.2 (2012-02-22)

  • Fix bug: file upload: file should always have a filename

0.2.1 (2012-02-22)

  • More flexible file upload

  • Rename fetch2 to request

  • Add auth parameter, instead of put basic authentication info in url

0.1.2 (2011-12-07)

  • Support basic auth

0.1 (2011-12-02)

  • First release