blob: 11f6f8942fc8997799130a69c32acf4f832a8649 [file] [log] [blame]
_ _ ____ _
___| | | | _ \| |
/ __| | | | |_) | |
| (__| |_| | _ <| |___
\___|\___/|_| \_\_____|
Things to do in project cURL. Please tell us what you think, contribute and
send us patches that improve things! Also check the
web section for various technical development notes.
All bugs documented in the KNOWN_BUGS document are subject for fixing!
* Introduce another callback interface for upload/download that makes one
less copy of data and thus a faster operation.
* More data sharing. curl_share_* functions already exist and work, and they
can be extended to share more. For example, enable sharing of the ares
channel and the connection cache.
* Introduce a new error code indicating authentication problems (for proxy
CONNECT error 407 for example). This cannot be an error code, we must not
return informational stuff as errors, consider a new info returned by
* Use 'struct lifreq' and SIOCGLIFADDR instead of 'struct ifreq' and
SIOCGIFADDR on newer Solaris versions as they claim the latter is obsolete.
To support ipv6 interface addresses properly.
* Add the following to curl_easy_getinfo(): GET_HTTP_IP, GET_FTP_IP and
GET_FTP_DATA_IP. Return a string with the used IP. Suggested by Alan.
* Add option that changes the interval in which the progress callback is
called at most.
* Make libcurl built with c-ares use c-ares' IPv6 abilities. They weren't
present when we first added c-ares support but they have been added since!
When this is done and works, we can actually start considering making c-ares
powered libcurl the default build (which of course would require that we'd
bundle the c-ares source code in the libcurl source code releases).
* Make the curl/*.h headers include the proper system includes based on what
was present at the time when configure was run. Currently, the sys/select.h
header is for example included by curl/multi.h only on specific platforms
we know MUST have it. This is error-prone. We therefore want the header
files to adapt to configure results. Those results must be stored in a new
header and they must use a curl name space, i.e not be HAVE_* prefix (as
that would risk collide with other apps that use libcurl and that runs
Work on this has been started but hasn't been finished, and the initial
patch and some details are found here:
LIBCURL - multi interface
* Make sure we don't ever loop because of non-blocking sockets return
EWOULDBLOCK or similar. The GnuTLS connection etc.
* Make transfers treated more carefully. We need a way to tell libcurl we
have data to write, as the current system expects us to upload data each
time the socket is writable and there is no way to say that we want to
upload data soon just not right now, without that aborting the upload. The
opposite situation should be possible as well, that we tell libcurl we're
ready to accept read data. Today libcurl feeds the data as soon as it is
available for reading, no matter what.
* Make curl_easy_perform() a wrapper-function that simply creates a multi
handle, adds the easy handle to it, runs curl_multi_perform() until the
transfer is done, then detach the easy handle, destroy the multi handle and
return the easy handle's return code. This will thus make everything
internally use and assume the multi interface. The select()-loop should use
* curl_multi_handle_control() - this can control the easy handle (while)
added to a multi handle in various ways:
o RESTART, unconditionally restart this easy handle's transfer from the
start, re-init the state
o RESTART_COMPLETED, restart this easy handle's transfer but only if the
existing transfer has already completed and it is in a "finished state".
o STOP, just stop this transfer and consider it completed
* More and better
* PRET is a command that primarily "drftpd" supports, which could be useful
when using libcurl against such a server. It is a non-standard and a rather
oddly designed command, but...
* When trying to connect passively to a server which only supports active
connections, libcurl returns CURLE_FTP_WEIRD_PASV_REPLY and closes the
connection. There could be a way to fallback to an active connection (and
vice versa).
* Make the detection of (bad) %0d and %0a codes in FTP url parts earlier in
the process to avoid doing a resolve and connect in vain.
* REST fix for servers not behaving well on >2GB requests. This should fail
if the server doesn't set the pointer to the requested index. The tricky
(impossible?) part is to figure out if the server did the right thing or
* Support the most common FTP proxies, Philip Newton provided a list
allegedly from ncftp:
* Make CURLOPT_FTPPORT support an additional port number on the IP/if/name,
like "blabla:[port]" or possibly even "blabla:[portfirst]-[portsecond]".
* FTP ASCII transfers do not follow RFC959. They don't convert the data
* Since USERPWD always override the user and password specified in URLs, we
might need another way to specify user+password for anonymous ftp logins.
* The FTP code should get a way of returning errors that is known to still
have the control connection alive and sound. Currently, a returned error
from within ftp-functions does not tell if the control connection is still
OK to use or not. This causes libcurl to fail to re-use connections
slightly too often.
* When doing CONNECT to a HTTP proxy, libcurl always uses HTTP/1.0. This has
never been reported as causing trouble to anyone, but should be considered
to use the HTTP version the user has chosen.
* "Better" support for persistent connections over HTTP 1.0
* Reading input (to send to the remote server) on stdin is a crappy solution
for library purposes. We need to invent a good way for the application to
be able to provide the data to send.
* Move the telnet support's network select() loop go away and merge the code
into the main transfer loop. Until this is done, the multi interface won't
work for telnet.
* Provide an option that allows for disabling specific SSL versions, such as
* Provide a libcurl API for setting mutex callbacks in the underlying SSL
library, so that the same application code can use mutex-locking
independently of OpenSSL or GnutTLS being used.
* Anton Fedorov's "dumpcert" patch:
* Evaluate/apply Gertjan van Wingerde's SSL patches:
* "Look at SSL cafile - quick traces look to me like these are done on every
request as well, when they should only be necessary once per ssl context
(or once per handle)". The major improvement we can rather easily do is to
make sure we don't create and kill a new SSL "context" for every request,
but instead make one for every connection and re-use that SSL context in
the same style connections are re-used. It will make us use slightly more
memory but it will libcurl do less creations and deletions of SSL contexts.
* Add an interface to libcurl that enables "session IDs" to get
exported/imported. Cris Bailiff said: "OpenSSL has functions which can
serialise the current SSL state to a buffer of your choice, and
recover/reset the state from such a buffer at a later date - this is used
by mod_ssl for apache to implement and SSL session ID cache".
* OpenSSL supports a callback for customised verification of the peer
certificate, but this doesn't seem to be exposed in the libcurl APIs. Could
it be? There's so much that could be done if it were! (brought by Chris
* Make curl's SSL layer capable of using other free SSL libraries. Such as
MatrixSSL (
* Peter Sylvester's patch for SRP on the TLS layer.
Awaits OpenSSL support for this, no need to support this in libcurl before
there's an OpenSSL release that does it.
* make the configure --with-ssl option first check for OpenSSL, then GnuTLS,
then NSS...
* Get NTLM working using the functions provided by libgcrypt, since GnuTLS
already depends on that to function. Not strictly SSL/TLS related, but
hey... Another option is to get available DES and MD4 source code from the
cryptopp library. They are fine license-wise, but are C++.
* SSL engine stuff?
* Work out a common method with Peter Sylvester's OpenSSL-patch for SRP
on the TLS to provide name and password
* Fix the connection phase to be non-blocking when multi interface is used
* Add a way to check if the connection seems to be alive, to correspond to
the SSL_peak() way we use with OpenSSL.
* Look over the implementation. The looping will have to "go away" from the
lib/ldap.c source file and get moved to the main network code so that the
multi interface and friends will work for LDAP as well.
* RTSP - RFC2326 (protocol - very HTTP-like, also contains URL description)
* RSYNC (no RFCs for protocol nor URI/URL format). An implementation should
most probably use an existing rsync library, such as librsync.
* Add option that is similar to -O but that takes the output file name from
the Content-Disposition: header, and/or uses the local file name used in
redirections for the cases the server bounces the request further to a
different file (name):
* "curl --sync[1-100].rss" or
"curl --sync{index,calendar,history}.html"
Downloads a range or set of URLs using the remote name, but only if the
remote file is newer than the local file. A Last-Modified HTTP date header
should also be used to set the mod date on the downloaded file.
(idea from "Brianiac")
* Globbing support for -d and -F, as in 'curl -d "name=foo[0-9]" URL'.
Requested by Dane Jensen and others. This is easily scripted though.
* Add an option that prevents cURL from overwriting existing local files. When
used, and there already is an existing file with the target file name
(either -O or -o), a number should be appended (and increased if already
existing). So that index.html becomes first index.html.1 and then
index.html.2 etc. Jeff Pohlmeyer suggested.
* "curl*.txt"
* The client could be told to use maximum N simultaneous parallel transfers
and then just make sure that happens. It should of course not make more
than one connection to the same remote host. This would require the client
to use the multi interface.
* Extending the capabilities of the multipart formposting. How about leaving
the ';type=foo' syntax as it is and adding an extra tag (headers) which
works like this: curl -F "coolfiles=@fil1.txt;headers=@fil1.hdr" where
fil1.hdr contains extra headers like
Content-Type: text/plain; charset=KOI8-R"
Content-Transfer-Encoding: base64
X-User-Comment: Please don't use browser specific HTML code
which should overwrite the program reasonable defaults (plain/text,
8bit...) (Idea brough to us by kromJx)
* ability to specify the classic computing suffixes on the range
specifications. For example, to download the first 500 Kilobytes of a file,
be able to specify the following for the -r option: "-r 0-500K" or for the
first 2 Megabytes of a file: "-r 0-2M". (Mark Smith suggested)
* --data-encode that URL encodes the data before posting (Kevin Roth suggested)
* Provide a way to make options bound to a specific URL among several on the
command line. Possibly by letting ':' separate options between URLs,
similar to this:
curl --data foo --url : \
--url : \
--url --data foo3
(More details:
The example would do a POST-GET-POST combination on a single command line.
* Consider extending 'roffit' to produce decent ASCII output, and use that
instead of (g)nroff when building src/hugehelp.c
* Make our own version of stunnel for simple port forwarding to enable HTTPS
and FTP-SSL tests without the stunnel dependency, and it could allow us to
provide test tools built with either OpenSSL or GnuTLS
* If perl wasn't found by the configure script, don't attempt to run the
tests but explain something nice why it doesn't.
* Extend the test suite to include more protocols. The telnet could just do
ftp or http operations (for which we have test servers).
* Make the test suite work on more platforms. OpenBSD and Mac OS. Remove
fork()s and it should become even more portable.
NEXT soname bump
* #undef CURL_FTP_HTTPSTYLE_HEAD in lib/ftp.c to remove the HTTP-style headers
from being output in NOBODY requests over ftp
* Combine some of the error codes to remove duplicates. The original
numbering should not be changed, and the old identifiers would be
macroed to the new ones in an CURL_NO_OLDIES section to help with
backward compatibility.
Candidates for removal and their replacements:
* curl_easy_cleanup() returns void, but curl_multi_cleanup() returns a
CURLMcode. These should be changed to be the same.
* remove obsolete defines from curl/curl.h
* make several functions use size_t instead of int in their APIs
* remove the following functions from the public API:
curl_mprintf (and variations)
They will instead become curlx_ - alternatives. That makes the curl app
still capable of building with them from source.
* Remove support for CURLOPT_FAILONERROR, it has gotten too kludgy and weird
internally. Let the app judge success or not for itself.