44 lines
1.7 KiB
Plaintext
44 lines
1.7 KiB
Plaintext
|
|
Known problems/bugs in neon -*- text -*-
|
|
---------------------------
|
|
|
|
* look at escaping logic again w.r.t. ?, # characters?
|
|
|
|
* 2818 requires that a on rejection of the SSL server cert, a "bad certificate"
|
|
message should be sent - this is not being done currently (and can probably
|
|
only be done with OpenSSL by actually doing cert verification in the verify
|
|
callback)
|
|
|
|
* ne_lock_discover does not handle multiple (shared) locks on
|
|
a single resource.
|
|
|
|
* ne_lock_refresh does not update the passed-in lock structure.
|
|
|
|
* SSL session caching issues; only cache for clean shutdowns, and
|
|
only cache on shutdown, since the SSL_SESSION may change during
|
|
an ne_session?
|
|
|
|
* what is passed as 'path' to req create hook: auth needs Request-URI;
|
|
how does that interact with proxies? also they will be passed NULL
|
|
for a CONNECT request, or "*" possibly as well.
|
|
|
|
* It would be nice to fail with a friendly error message if a client
|
|
cert is requested by the srever but one is not provided. Currently,
|
|
returning -1 from the provide_client_cert function would allow that
|
|
(as it forces the SSL handshake to fail), but that would prevent
|
|
opportunistic use of client certificates, of the "SSLVerifyClient
|
|
optional" variety.
|
|
|
|
* D.J. Heap has a proxy which returns a 401 in response to a CONNECT;
|
|
relax the ne_auth rules to allow this since it's unambiguous.
|
|
|
|
* Error handling from ne__pull_request_body/send_request_body is
|
|
buggy: socket errors are not distinguished from body provider errors;
|
|
the connection must be closed in ne_request after a body provider
|
|
error.
|
|
|
|
* Check whether the following always return UTF-8-encoded strings:
|
|
- ne_ssl_clicert_name
|
|
- ne_ssl_cert_identity
|
|
|