Skip to content

gh-135056: Add a --header CLI argument to http.server#135057

Open
aisipos wants to merge 47 commits intopython:mainfrom
aisipos:https-server-cors-issue-135056
Open

gh-135056: Add a --header CLI argument to http.server#135057
aisipos wants to merge 47 commits intopython:mainfrom
aisipos:https-server-cors-issue-135056

Conversation

@aisipos
Copy link
Copy Markdown

@aisipos aisipos commented Jun 3, 2025

As proposed in #135056, Add a --cors command line argument to the stdlib http.server module, which will add an Access-Control-Allow-Origin: * header to all responses.

Invocation:

python -m http.server --cors

As part of this implementation, add a response_headers argument to SimpleHTTPRequestHandler and HTTPServer, which allows callers to add addition headers to the response. Ideally it would have been possible to just have made a CorsHttpServer class, but a couple of issues made that difficult:

  • The http.server CLI uses more than one HTTP Server class, in order to support TLS/HTTPS. So a single CorsHttpServer child class wouldn't work to support both use cases.
  • Much of the work in specifying HTTP behavior is handled by the various RequestHandler classes. However, the HttpServer classes didn't have an easy way to pass arguments down into the instantiated handlers.

As a result, this PR updates both HTTPServer and SimpleHTTPRequestHandler to accept a response_headers argument, which allows callers to specify an additional set of HTTP headers to pass in the response.

  • HTTPServer now overrides finish_request to pass this new kwarg down to its RequestHandler.
  • SimpleHTTPRequestHandler now accepts a resposnse_headers kwarg, to optionally specify a dictionary of additional headers to send in the response.

Care is taken to not pass the response_headers argument to any instance constructors when not provided, to ensure backwards compatibility. I tried to keep the implementation as short and simple as possible.

With the addition of a response_headers argument, we allow ourselves to have a future possible custom header http argument, such as:

python -m http.server -H 'custom-header: custom-value'

📚 Documentation preview 📚: https://cpython-previews--135057.org.readthedocs.build/

@python-cla-bot
Copy link
Copy Markdown

python-cla-bot bot commented Jun 3, 2025

All commit authors signed the Contributor License Agreement.

CLA signed

@bedevere-app
Copy link
Copy Markdown

bedevere-app bot commented Jun 3, 2025

Most changes to Python require a NEWS entry. Add one using the blurb_it web app or the blurb command-line tool.

If this change has little impact on Python users, wait for a maintainer to apply the skip news label instead.

Add a --cors command line argument to the stdlib http.server module, which will
add an `Access-Control-Allow-Origin: *` header to all responses. As part of this
implementation, add a `response_headers` argument to SimpleHTTPRequestHandler
and HttpServer, which allows callers to add addition headers to the response.
@aisipos aisipos force-pushed the https-server-cors-issue-135056 branch from 3f11652 to 0d02fbe Compare June 3, 2025 05:24
Copy link
Copy Markdown
Member

@hugovk hugovk left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

(I'd prefer a general headers option, but will comment on the issue or Discourse topic)

@aisipos
Copy link
Copy Markdown
Author

aisipos commented Jun 3, 2025

test_wsgiref is failing. I'll look into it.

This fixes the breakage to HttpServer as used by wsgiref.
@aisipos
Copy link
Copy Markdown
Author

aisipos commented Jun 3, 2025

test_wsgiref fixed in a3256fd. This should fix any backwards incompatibility errors erroneously introduced in the first commit.

@donbarbos
Copy link
Copy Markdown
Contributor

I think it's worth adding to this What's New entry (./Doc/whatsnew/3.15.rst)

@Zheaoli
Copy link
Copy Markdown
Contributor

Zheaoli commented Jun 6, 2025

For me, I don't think add --cors CLI argument would be a good idea. Base on following reasons:

  1. The CORS policy is a complicated idea. Six response headers are included by the word(If I'm correct). If you set the Access-Control-Allow-Origin, and now the people may need Access-Control-Allow-Methods( Allow for OPTION method). What is the next argument we need to add?
  2. The CLI for http.server is just for a debug target. So we design the CLI as simple as we can. The developer don't need to care about any extra detail when they run just a simple debug server.
  3. if we need CORS policy in the future. I suggest we setup all the header for developer and don't need to add cli for it.

@aisipos aisipos requested a review from hugovk June 6, 2025 23:05
@picnixz picnixz self-requested a review June 7, 2025 23:15
@hugovk
Copy link
Copy Markdown
Member

hugovk commented Jun 9, 2025

@hugovk
Copy link
Copy Markdown
Member

hugovk commented Jun 9, 2025

(I'd prefer a general headers option, but will comment on the issue or Discourse topic)

https://discuss.python.org/t/any-interest-in-adding-a-cors-option-to-python-m-http-server/92120/24

Copy link
Copy Markdown
Member

@picnixz picnixz left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'm not very fond of how the HTTP server class is growing more and more with more __init__ parameters, but I don't have a better idea for now. Maybe a generic configuration object but this would be an overkill for this class in particular I think.


.. versionadded:: 3.14

.. option:: --cors
Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

As Hugo said, since we're anyway exposing response-headers, I think we should also expose it from the CLI. It could be useful for users in general (e.g., --add-header NAME VALUE with the -H alias).

@@ -0,0 +1,2 @@
Add a ``--cors`` cli option to :program:`python -m http.server`. Contributed by
Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Let's also update What's New/3.15.rst

Copy link
Copy Markdown
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I used blurb to make this entry in NEWS.d, not knowing when it's appropriate to edit the main 3.15.rst file. I think once we know if we're doing --cors / --header , or both, I can make the appropriate update to What's New/3.15.rst

def __init__(self, server_address, RequestHandlerClass,
bind_and_activate=True, *, certfile, keyfile=None,
password=None, alpn_protocols=None):
password=None, alpn_protocols=None, response_headers=None):
Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
password=None, alpn_protocols=None, response_headers=None):
password=None, alpn_protocols=None, **http_server_kwargs):

And pass http_server_kwargs to super()

Comment on lines +133 to +138
args = (request, client_address, self)
kwargs = {}
response_headers = getattr(self, 'response_headers', None)
if response_headers:
kwargs['response_headers'] = self.response_headers
self.RequestHandlerClass(*args, **kwargs)
Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
args = (request, client_address, self)
kwargs = {}
response_headers = getattr(self, 'response_headers', None)
if response_headers:
kwargs['response_headers'] = self.response_headers
self.RequestHandlerClass(*args, **kwargs)
kwargs = {}
if hasattr(self, 'response_headers'):
kwargs['response_headers'] = self.response_headers
self.RequestHandlerClass(request, client_address, self, **kwargs)

Copy link
Copy Markdown
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@picnixz I made this requested change in 77b5fff. Note though that now HTTPServer will pass response_headers to the RequestHandler class even if response_headers is None or {}. Most RequestHandler implementation constructor implementations don't take this argument, so in order for this to work I added **kwargs as an argument to BaseRequestHandler.__init__. My earlier implementation was trying to prevent this, to keep any changes to only http/server.py and not need to touch anything in socketserver.py. Perhaps the **kwargs addition is ok, or I'm open to other solutions if we can think of better ones.

Comment on lines +708 to +712
def __init__(self, *args, directory=None, response_headers=None, **kwargs):
if directory is None:
directory = os.getcwd()
self.directory = os.fspath(directory)
self.response_headers = response_headers or {}
Copy link
Copy Markdown
Member

@picnixz picnixz Jun 9, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
def __init__(self, *args, directory=None, response_headers=None, **kwargs):
if directory is None:
directory = os.getcwd()
self.directory = os.fspath(directory)
self.response_headers = response_headers or {}
def __init__(self, *args, directory=None, response_headers=None, **kwargs):
if directory is None:
directory = os.getcwd()
self.directory = os.fspath(directory)
self.response_headers = response_headers

You're already checking for is not None later

ServerClass=ThreadingHTTPServer,
protocol="HTTP/1.0", port=8000, bind=None,
tls_cert=None, tls_key=None, tls_password=None):
tls_cert=None, tls_key=None, tls_password=None, response_headers=None):
Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
tls_cert=None, tls_key=None, tls_password=None, response_headers=None):
tls_cert=None, tls_key=None, tls_password=None,
response_headers=None):

Let's group the parameters per purpose

Comment on lines +1078 to +1082
handler_args = (request, client_address, self)
handler_kwargs = dict(directory=args.directory)
if self.response_headers:
handler_kwargs['response_headers'] = self.response_headers
self.RequestHandlerClass(*handler_args, **handler_kwargs)
Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
handler_args = (request, client_address, self)
handler_kwargs = dict(directory=args.directory)
if self.response_headers:
handler_kwargs['response_headers'] = self.response_headers
self.RequestHandlerClass(*handler_args, **handler_kwargs)
self.RequestHandlerClass(request, client_address, self,
directory=args.directory,
response_headers=self.response_headers)

)
else:
self.server = HTTPServer(('localhost', 0), self.request_handler)
self.server = HTTPServer(('localhost', 0), self.request_handler,
Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

You must also modify create_https_server appropriately

Comment on lines +832 to +834
server_kwargs = dict(
response_headers = {'Access-Control-Allow-Origin': '*'}
)
Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
server_kwargs = dict(
response_headers = {'Access-Control-Allow-Origin': '*'}
)
server_kwargs = {
'response_headers': {'Access-Control-Allow-Origin': '*'}
}

server_kwargs = dict(
response_headers = {'Access-Control-Allow-Origin': '*'}
)
def test_cors(self):
Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
def test_cors(self):
def test_cors(self):

@aisipos aisipos force-pushed the https-server-cors-issue-135056 branch from 3024d3d to 5f89c97 Compare June 20, 2025 01:50
@aisipos
Copy link
Copy Markdown
Author

aisipos commented Jun 20, 2025

@picnixz I have made all your suggested changes in 77b5fff . I have also implemented a generic -H or --header cli argument in 5f89c97. Now this PR contains both --cors and --header. I don't know if we want both, there doesn't seem to yet be consensus on what we prefer, although Core devs so far seem to lean on just --header.

@hugovk
Copy link
Copy Markdown
Member

hugovk commented Jun 20, 2025

I think we should just have --header, as that unlocks the ability to enable CORS. We can still add --cors later if there's demand and consensus.

@hugovk
Copy link
Copy Markdown
Member

hugovk commented Jun 20, 2025

And what are your thoughts on positional args like HTTPie?

https://discuss.python.org/t/any-interest-in-adding-a-cors-option-to-python-m-http-server/92120/24

Support of the ``'If-Modified-Since'`` header.

.. versionchanged:: next
Support ``response_headers`` as an instance argument.
Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Isn’t this redundant with the entry already under the constructor heading?

Copy link
Copy Markdown
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Perhaps - it seems the constructor documentation is used to make a brief mention of each argument and when it was added, with more detail being filled in later. My latest commits make several changes requested elsewhere for other reasons, but if the current version is still too redundant in multiple places I can make some more edits.

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I don't think we should add this information. There is no notion of an "instance argument": it should rather be an instance attribute, and this should be documented through a .. attribute::, below .. attribute:: extensions_map

self.RequestHandlerClass(request, client_address, self,
directory=args.directory)
directory=args.directory,
response_headers=self.response_headers)
Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Why not do this the same way the --directory or --protocol options are implemented? Either way should avoid adding internal parameters to unrelated HTTPServer and BaseRequestHandler classes.

You could build the response_headers dictionary before the DualStackServerMixin class, and then pass it by referencing the outer scope like is already done with args.directory:

Suggested change
response_headers=self.response_headers)
response_headers=response_headers)

Or set the response_headers attribute on the SimpleHTTPRequestHandler class rather than in its constructor, like is done for protocol_verison in the test function.

Copy link
Copy Markdown
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I followed your advice, see b1026d2. I made response_headers an argument to SimpleHTTPRequestHandler only, and send the argument to it in the DualStackServerMixin class.

The *directory* parameter accepts a :term:`path-like object`.

.. versionchanged:: next
The *response_headers* parameter accepts an optional dictionary of
Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

In previous versions, this was not a valid parameter at all.

Suggested change
The *response_headers* parameter accepts an optional dictionary of
Added *response_headers*, which accepts an optional dictionary of

Also, did you consider accepting a list or iterable of (name, value) pairs instead, like returned by http.client.HTTPResponse.getheaders? That would be better for sending multiple Set-Cookie fields.

Copy link
Copy Markdown
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Ah yes, sending multiple headers of the same name would indeed be necessary. I updated to use an iterable of name value pairs instead in 7a793f2


.. versionchanged:: next
The *response_headers* parameter accepts an optional dictionary of
additional HTTP headers to add to each response.
Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Might be worth clarifying how these fields interact with other fields such as Server specified under BaseHTTPRequestHandler.send_response, and Last-Modified under do_GET.

Also clarify which responses the fields are included in, assuming it wasn’t your intention to include them for 404 Not Found, 304 Not Modified, lower-level errors, etc.

Copy link
Copy Markdown
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

In the latest commits, I've noted that the custom headers are only sent in success cases. What do you mean by interaction though? The custom headers currently get sent after Last-Modified, should I mention the placement of the custom headers and that they appear after Last-Modified?

Comment on lines +754 to +756
if self.response_headers is not None:
for header, value in self.response_headers.items():
self.send_header(header, value)
Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Or is moving this to an extended send_response override an option? That way you would include the fields for all responses.

if directory is None:
directory = os.getcwd()
self.directory = os.fspath(directory)
self.response_headers = response_headers
Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Clarify as an internal private attribute:

Suggested change
self.response_headers = response_headers
self._response_headers = response_headers

Or document SimpleHTTPRequestHandler.response_headers as a public attribute.

Copy link
Copy Markdown
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

So far I have intended response_headers to be a public attribute. Do you mean add documentation to the docstring of SimpleHTTPRequestHandler or more documentation in Doc/library/http.server.rst?

else:
self.server = HTTPServer(('localhost', 0), self.request_handler)
self.server = HTTPServer(('localhost', 0), self.request_handler,
**self.server_kwargs)
Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This appears to only be testing the undocumented or internal HTTPServer parameter. It would be good to test the new documented SimpleHTTPRequestHandler parameter instead or as well.

Copy link
Copy Markdown
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I have removed server_kwargs in the latest updates, and updated the tests. The only external change now is response_headers as an instance arg to SimpleHTTPRequestHandler

@aisipos aisipos requested a review from vadmium July 10, 2025 07:22
@aisipos
Copy link
Copy Markdown
Author

aisipos commented Oct 10, 2025

There are some new build breakages, I'll investigate and fix.

@aisipos
Copy link
Copy Markdown
Author

aisipos commented Oct 10, 2025

There are some new build breakages, I'll investigate and fix.

Build breaks fixed in ed0b0b3 and 79c577b

@aisipos aisipos requested a review from picnixz October 10, 2025 06:23
into https-server-cors-issue-135056
@aisipos
Copy link
Copy Markdown
Author

aisipos commented Dec 12, 2025

@picnixz I have merged master into the PR and fixed the conflict in the what's new file - would you be able to give this another review?

Copy link
Copy Markdown
Member

@hugovk hugovk left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks!

@picnixz Any final comments?

@picnixz
Copy link
Copy Markdown
Member

picnixz commented Jan 17, 2026

I would like to review it when I am back at home so it will need to wait like 10 days or so!

@aisipos
Copy link
Copy Markdown
Author

aisipos commented Mar 19, 2026

@picnixz Thanks for your help in reviewing this PR. It would be nice to get it merged before the 3.15 beta - do you have any further thoughts on this PR?

@hugovk
Copy link
Copy Markdown
Member

hugovk commented Mar 21, 2026

@aisipos Just in case @picnixz doesn't have time to re-check, please ping me towards the end of April and we can merge if there's no blocking objections. We can always bugfix during the beta if needed.

Copy link
Copy Markdown
Member

@picnixz picnixz left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

We just need one additional test where you re-specify the Content-Type and Content-Length headers. I wonder whether we should first send the extra headers and only send those two if they are not in the extra headers or just reject those two as being customizable.

Otherwise I think I am ok with this approach. It is generic enough that we do not to worry too much. The question is whether we should consider those extra headers as "safe" or not and whether validation is necessary.

From a release perspective @hugovk : is it ok to remove/drastically redesign a feature introduced in alpha if it was not that ready?

HTTP responses.
(Contributed by Anton I. Sipos in :gh:`135057`.)

* Added a ``-H`` or ``--header`` flag to the :program:`python -m http.server`
Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
* Added a ``-H`` or ``--header`` flag to the :program:`python -m http.server`
* Added the ``-H/--header`` option to the :program:`python -m http.server`

f.close()

def _send_extra_response_headers(self):
"""Send the headers stored in self.extra_response_headers"""
Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
"""Send the headers stored in self.extra_response_headers"""
"""Send the headers stored in self.extra_response_headers."""

@hugovk
Copy link
Copy Markdown
Member

hugovk commented Mar 21, 2026

From a release perspective @hugovk : is it ok to remove/drastically redesign a feature introduced in alpha if it was not that ready?

Things can be reverted right through to end of RC (see the incremental GC in 3.13 for an extreme case).

As we get closer to final release, each stage is intended to increase the stability, so I'd say no to drastic redesigns, and other changes should balance the risk/benefit and not introduce new features.

@picnixz
Copy link
Copy Markdown
Member

picnixz commented Mar 21, 2026

So apart from:

  • Tests when we redefine headers.
  • Decide what to do with redefinitions.
  • Decide what to do with validation (cc @sethmlarson)

I am ok with this PR. I would say that extra headers should be by-passed and should not be checked against existing ones (possibly sending twice the Content-Type header) unless the RFC forbids sending headers twice (in which case we need to use user-defined ones).

For validation, I would suggest we do nothing and add a security warning. Since http.server is not meant to be used in production it should be fine to have untrusted headers here.

@sethmlarson
Copy link
Copy Markdown
Contributor

We can validate for CR/LF since those aren't allowed in HTTP headers, beyond that I think if it's set then it's okay to send values that are supplied to users. I'm concerned about being able to set duplicate headers for Content-Type, maybe we should disallow this at the CLI level or accept the user-supplied value?

@picnixz
Copy link
Copy Markdown
Member

picnixz commented Mar 25, 2026

I'm concerned about being able to set duplicate headers for Content-Type, maybe we should disallow this at the CLI level or accept the user-supplied value?

As http.server is really not meant for production, I really don't want to provide the same level of security guarantees as http.client. Using the user-defined headers for the headers we supply is fine but it's a bit weird because those headers will be always be sent back whatever the response is. So ignoring them is also fine.

How about rejecting the headers that we internally supply for now and maybe extend the functionality if needs arise?

@sethmlarson
Copy link
Copy Markdown
Contributor

@picnixz That would work for me! :)

@picnixz
Copy link
Copy Markdown
Member

picnixz commented Mar 25, 2026

Ok, so to summarize @aisipos:

  • Please make it so that we explicitly reject the headers that we already supply.
  • Update the documentation about those headers.
  • Add tests.

@aisipos aisipos force-pushed the https-server-cors-issue-135056 branch from 43b0d51 to 3a4fed6 Compare April 12, 2026 21:55
@aisipos
Copy link
Copy Markdown
Author

aisipos commented Apr 12, 2026

@picnixz OK, thanks for your help. In the latest 2 commits I updated the code to not overwrite headers that the server has already sent (including a new test for this), and updated the documentation to describe this. It's ready for another re-review.

@picnixz
Copy link
Copy Markdown
Member

picnixz commented Apr 12, 2026

Please fix the tests first.

@aisipos
Copy link
Copy Markdown
Author

aisipos commented Apr 12, 2026

@picnixz OK sorry, I made the new attribute access I added more resilient to mocking in tests. There are still a couple of test runs failing on windows profiling, but they seem unrelated?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

8 participants