5 \___|\___/|_| \_\_____|
7 Things that could be nice to do in the future
9 Things to do in project curl. Please tell us what you think, contribute and
10 send us patches that improve things!
12 Be aware that these are things that we could do, or have once been considered
13 things we could do. If you want to work on any of these areas, please
14 consider bringing it up for discussions first on the mailing list so that we
15 all agree it is still a good idea for the project!
17 All bugs documented in the KNOWN_BUGS document are subject for fixing!
20 1.1 Option to refuse usernames in URLs
23 1.4 signal-based resolver timeouts
24 1.5 get rid of PATH_MAX
25 1.6 Modified buffer size approach
26 1.7 Support HTTP/2 for HTTP(S) proxies
27 1.8 CURLOPT_RESOLVE for any port number
28 1.9 Cache negative name resolves
29 1.10 auto-detect proxy
30 1.11 minimize dependencies with dynamically loaded modules
31 1.12 updated DNS server while running
33 1.14 Typesafe curl_easy_setopt()
34 1.15 Monitor connections in the connection pool
35 1.16 Try to URL encode given URL
36 1.17 Add support for IRIs
37 1.18 try next proxy if one doesn't work
38 1.19 Timeout idle connections from the pool
39 1.20 SRV and URI DNS records
40 1.21 API for URL parsing/splitting
41 1.23 Offer API to flush the connection pool
42 1.24 TCP Fast Open for windows
43 1.25 Expose tried IP addresses that failed
44 1.26 CURL_REFUSE_CLEARTEXT
45 1.27 hardcode the "localhost" addresses
47 2. libcurl - multi interface
49 2.2 Better support for same name resolves
50 2.3 Non-blocking curl_multi_remove_handle()
51 2.4 Split connect and authentication process
52 2.5 Edge-triggered sockets should work
55 3.2 Provide cmake config-file
59 4.2 Alter passive/active on failure and retry
60 4.3 Earlier bad letter detection
61 4.4 REST for large files
63 4.6 GSSAPI via Windows SSPI
64 4.7 STAT for LIST without data connection
65 4.8 Option to ignore private IP addresses in PASV response
68 5.1 Better persistency for HTTP 1.0
69 5.2 support FF3 sqlite cookie files
70 5.3 Rearrange request header order
72 5.6 Refuse "downgrade" redirects
74 5.8 Leave secure cookies alone
78 6.2 ditch telnet-specific select
79 6.3 feature negotiation debug data
83 7.2 Enhanced capability support
84 7.3 Add CURLOPT_MAIL_CLIENT option
88 8.2 Enhanced capability support
91 9.1 Enhanced capability support
94 10.1 SASL based authentication mechanisms
97 11.1 File listing support
98 11.2 Honor file timestamps
100 11.4 Create remote directories
106 13.1 Disable specific versions
107 13.2 Provide mutex locking API
108 13.3 Support in-memory certs/ca certs/keys
109 13.4 Cache/share OpenSSL contexts
110 13.5 Export session ids
111 13.6 Provide callback for cert verification
112 13.7 improve configure --with-ssl
114 13.11 Support intermediate & root pinning for PINNEDPUBLICKEY
119 14.1 SSL engine stuff
120 14.2 check connection
123 15.1 Add support for client certificate authentication
124 15.2 Add support for custom server certificate validation
125 15.3 Add support for the --ciphers option
128 16.1 Other authentication mechanisms
129 16.2 Add QOP support to GSSAPI authentication
130 16.3 Support binary messages (i.e.: non-base64)
134 17.2 SFTP performance
135 17.3 Support better than MD5 hostkey hash
136 17.4 Support CURLOPT_PREQUOTE
138 18. Command line tool
141 18.3 prevent file overwriting
142 18.4 simultaneous parallel transfers
143 18.5 UTF-8 filenames in Content-Disposition
144 18.6 warning when setting an option
145 18.7 warning if curl version is not in sync with libcurl version
146 18.8 offer color-coded HTTP header output
147 18.9 Choose the name of file in braces for complex URLs
148 18.10 improve how curl works in a windows console window
149 18.11 -w output to stderr
150 18.12 keep running, read instructions from pipe/socket
151 18.13 support metalink in http headers
152 18.14 --fail without --location should treat 3xx as a failure
153 18.15 --retry should resume
154 18.16 send only part of --data
155 18.17 consider file name from the redirected URL with -O ?
156 18.18 retry on network is unreachable
160 19.2 Enable PIE and RELRO by default
164 20.2 nicer lacking perl message
165 20.3 more protocols supported
166 20.4 more platforms supported
167 20.5 Add support for concurrent connections
168 20.6 Use the RFC6265 test suite
171 21.1 http-style HEAD output for FTP
172 21.2 combine error codes
173 21.3 extend CURLOPT_SOCKOPTFUNCTION prototype
175 22. Next major release
176 22.1 cleanup return codes
177 22.2 remove obsolete defines
179 22.4 remove several functions
180 22.5 remove CURLOPT_FAILONERROR
181 22.6 remove CURLOPT_DNS_USE_GLOBAL_CACHE
182 22.7 remove progress meter from libcurl
183 22.8 remove 'curl_httppost' from public
185 ==============================================================================
189 1.1 Option to refuse usernames in URLs
191 There's a certain risk for application in allowing user names in URLs. For
192 example: if the wrong person gets to set the URL and manages to set a user
193 name in there when .netrc is used, the application may send along a password
194 that otherwise the person couldn't provide.
196 A new libcurl option could be added to allow applications to switch off this
197 feature and thus avoid a potential risk.
199 1.2 More data sharing
201 curl_share_* functions already exist and work, and they can be extended to
202 share more. For example, enable sharing of the ares channel and the
207 Use 'struct lifreq' and SIOCGLIFADDR instead of 'struct ifreq' and
208 SIOCGIFADDR on newer Solaris versions as they claim the latter is obsolete.
209 To support IPv6 interface addresses for network interfaces properly.
211 1.4 signal-based resolver timeouts
213 libcurl built without an asynchronous resolver library uses alarm() to time
214 out DNS lookups. When a timeout occurs, this causes libcurl to jump from the
215 signal handler back into the library with a sigsetjmp, which effectively
216 causes libcurl to continue running within the signal handler. This is
217 non-portable and could cause problems on some platforms. A discussion on the
218 problem is available at https://curl.haxx.se/mail/lib-2008-09/0197.html
220 Also, alarm() provides timeout resolution only to the nearest second. alarm
221 ought to be replaced by setitimer on systems that support it.
223 1.5 get rid of PATH_MAX
225 Having code use and rely on PATH_MAX is not nice:
226 https://insanecoding.blogspot.com/2007/11/pathmax-simply-isnt.html
228 Currently the SSH based code uses it a bit, but to remove PATH_MAX from there
229 we need libssh2 to properly tell us when we pass in a too small buffer and
230 its current API (as of libssh2 1.2.7) doesn't.
232 1.6 Modified buffer size approach
234 Current libcurl allocates a fixed 16K size buffer for download and an
235 additional 16K for upload. They are always unconditionally part of the easy
236 handle. If CRLF translations are requested, an additional 32K "scratch
237 buffer" is allocated. A total of 64K transfer buffers in the worst case.
239 First, while the handles are not actually in use these buffers could be freed
240 so that lingering handles just kept in queues or whatever waste less memory.
242 Secondly, SFTP is a protocol that needs to handle many ~30K blocks at once
243 since each need to be individually acked and therefore libssh2 must be
244 allowed to send (or receive) many separate ones in parallel to achieve high
245 transfer speeds. A current libcurl build with a 16K buffer makes that
246 impossible, but one with a 512K buffer will reach MUCH faster transfers. But
247 allocating 512K unconditionally for all buffers just in case they would like
248 to do fast SFTP transfers at some point is not a good solution either.
250 Dynamically allocate buffer size depending on protocol in use in combination
251 with freeing it after each individual transfer? Other suggestions?
253 1.7 Support HTTP/2 for HTTP(S) proxies
255 Support for doing HTTP/2 to HTTP and HTTPS proxies is still missing.
257 1.8 CURLOPT_RESOLVE for any port number
259 This option allows applications to set a replacement IP address for a given
260 host + port pair. Consider making support for providing a replacement address
261 for the host name on all port numbers.
263 See https://github.com/curl/curl/issues/1264
265 1.9 Cache negative name resolves
267 A name resolve that has failed is likely to fail when made again within a
268 short period of time. Currently we only cache positive responses.
270 1.10 auto-detect proxy
272 libcurl could be made to detect the system proxy setup automatically and use
273 that. On Windows, macOS and Linux desktops for example.
275 The pull-request to use libproxy for this was deferred due to doubts on the
276 reliability of the dependency and how to use it:
277 https://github.com/curl/curl/pull/977
279 libdetectproxy is a (C++) library for detecting the proxy on Windows
280 https://github.com/paulharris/libdetectproxy
282 1.11 minimize dependencies with dynamically loaded modules
284 We can create a system with loadable modules/plug-ins, where these modules
285 would be the ones that link to 3rd party libs. That would allow us to avoid
286 having to load ALL dependencies since only the necessary ones for this
287 app/invoke/used protocols would be necessary to load. See
288 https://github.com/curl/curl/issues/349
290 1.12 updated DNS server while running
292 If /etc/resolv.conf gets updated while a program using libcurl is running, it
293 is may cause name resolves to fail unless res_init() is called. We should
294 consider calling res_init() + retry once unconditionally on all name resolve
295 failures to mitigate against this. Firefox works like that. Note that Windows
296 doesn't have res_init() or an alternative.
298 https://github.com/curl/curl/issues/2251
302 By adding support for DNS-over-HTTPS curl could resolve host names using a
303 totally separate name server than the standard system resolver, while at the
304 same time doing so over a communication channel that enhances privacy and
307 https://github.com/curl/curl/wiki/DNS-over-HTTPS
309 1.14 Typesafe curl_easy_setopt()
311 One of the most common problems in libcurl using applications is the lack of
312 type checks for curl_easy_setopt() which happens because it accepts varargs
313 and thus can take any type.
315 One possible solution to this is to introduce a few different versions of the
316 setopt version for the different kinds of data you can set.
318 curl_easy_set_num() - sets a long value
320 curl_easy_set_large() - sets a curl_off_t value
322 curl_easy_set_ptr() - sets a pointer
324 curl_easy_set_cb() - sets a callback PLUS its callback data
326 1.15 Monitor connections in the connection pool
328 libcurl's connection cache or pool holds a number of open connections for the
329 purpose of possible subsequent connection reuse. It may contain a few up to a
330 significant amount of connections. Currently, libcurl leaves all connections
331 as they are and first when a connection is iterated over for matching or
332 reuse purpose it is verified that it is still alive.
334 Those connections may get closed by the server side for idleness or they may
335 get a HTTP/2 ping from the peer to verify that they're still alive. By adding
336 monitoring of the connections while in the pool, libcurl can detect dead
337 connections (and close them) better and earlier, and it can handle HTTP/2
338 pings to keep such ones alive even when not actively doing transfers on them.
340 1.16 Try to URL encode given URL
342 Given a URL that for example contains spaces, libcurl could have an option
343 that would try somewhat harder than it does now and convert spaces to %20 and
344 perhaps URL encoded byte values over 128 etc (basically do what the redirect
345 following code already does).
347 https://github.com/curl/curl/issues/514
349 1.17 Add support for IRIs
351 IRIs (RFC 3987) allow localized, non-ascii, names in the URL. To properly
352 support this, curl/libcurl would need to translate/encode the given input
353 from the input string encoding into percent encoded output "over the wire".
355 To make that work smoothly for curl users even on Windows, curl would
356 probably need to be able to convert from several input encodings.
358 1.18 try next proxy if one doesn't work
360 Allow an application to specify a list of proxies to try, and failing to
361 connect to the first go on and try the next instead until the list is
362 exhausted. Browsers support this feature at least when they specify proxies
365 https://github.com/curl/curl/issues/896
367 1.19 Timeout idle connections from the pool
369 libcurl currently keeps connections in its connection pool for an indefinite
370 period of time, until it either gets reused, gets noticed that it has been
371 closed by the server or gets pruned to make room for a new connection.
373 To reduce overhead (especially for when we add monitoring of the connections
374 in the pool), we should introduce a timeout so that connections that have
375 been idle for N seconds get closed.
377 1.20 SRV and URI DNS records
379 Offer support for resolving SRV and URI DNS records for libcurl to know which
380 server to connect to for various protocols (including HTTP!).
382 1.21 API for URL parsing/splitting
384 libcurl has always parsed URLs internally and never exposed any API or
385 features to allow applications to do it. Still most or many applications
386 using libcurl need that ability. In polls to users, we've learned that many
387 libcurl users would like to see and use such an API.
389 1.23 Offer API to flush the connection pool
391 Sometimes applications want to flush all the existing connections kept alive.
392 An API could allow a forced flush or just a forced loop that would properly
393 close all connections that have been closed by the server already.
395 1.24 TCP Fast Open for windows
397 libcurl supports the CURLOPT_TCP_FASTOPEN option since 7.49.0 for Linux and
398 Mac OS. Windows supports TCP Fast Open starting with Windows 10, version 1607
399 and we should add support for it.
401 1.25 Expose tried IP addresses that failed
403 When libcurl fails to connect to a host, it should be able to offer the
404 application the list of IP addresses that were used in the attempt.
406 https://github.com/curl/curl/issues/2126
408 1.26 CURL_REFUSE_CLEARTEXT
410 An environment variable that when set will make libcurl refuse to use any
411 cleartext network protocol. That's all non-encrypted ones (FTP, HTTP, Gopher,
412 etc). By adding the check to libcurl and not just curl, this environment
413 variable can then help users to block all libcurl-using programs from
414 accessing the network using unsafe protocols.
416 The variable could be given some sort of syntax or different levels and be
417 used to also allow for example users to refuse libcurl to do transfers with
418 HTTPS certificate checks disabled.
420 It could also offer to refuse usernames in URLs (see TODO 1.1)
422 1.27 hardcode the "localhost" addresses
424 There's this new spec getting adopted that says "localhost" should always and
425 unconditionally be a local address and not get resolved by a DNS server. A
426 fine way for curl to fix this would be to simply hard-code the response to
427 127.0.0.1 and/or ::1 (depending on what IP versions that are requested). This
428 is what the browsers probably will do with this hostname.
430 https://bugzilla.mozilla.org/show_bug.cgi?id=1220810
432 https://tools.ietf.org/html/draft-ietf-dnsop-let-localhost-be-localhost-02
434 2. libcurl - multi interface
436 2.1 More non-blocking
438 Make sure we don't ever loop because of non-blocking sockets returning
439 EWOULDBLOCK or similar. Blocking cases include:
441 - Name resolves on non-windows unless c-ares or the threaded resolver is used
442 - SOCKS proxy handshakes
445 - The "DONE" operation (post transfer protocol-specific actions) for the
446 protocols SFTP, SMTP, FTP. Fixing Curl_done() for this is a worthy task.
448 2.2 Better support for same name resolves
450 If a name resolve has been initiated for name NN and a second easy handle
451 wants to resolve that name as well, make it wait for the first resolve to end
452 up in the cache instead of doing a second separate resolve. This is
453 especially needed when adding many simultaneous handles using the same host
454 name when the DNS resolver can get flooded.
456 2.3 Non-blocking curl_multi_remove_handle()
458 The multi interface has a few API calls that assume a blocking behavior, like
459 add_handle() and remove_handle() which limits what we can do internally. The
460 multi API need to be moved even more into a single function that "drives"
461 everything in a non-blocking manner and signals when something is done. A
462 remove or add would then only ask for the action to get started and then
463 multi_perform() etc still be called until the add/remove is completed.
465 2.4 Split connect and authentication process
467 The multi interface treats the authentication process as part of the connect
468 phase. As such any failures during authentication won't trigger the relevant
469 QUIT or LOGOFF for protocols such as IMAP, POP3 and SMTP.
471 2.5 Edge-triggered sockets should work
473 The multi_socket API should work with edge-triggered socket events. One of
474 the internal actions that need to be improved for this to work perfectly is
475 the 'maxloops' handling in transfer.c:readwrite_data().
479 3.2 Provide cmake config-file
481 A config-file package is a set of files provided by us to allow applications
482 to write cmake scripts to find and use libcurl easier. See
483 https://github.com/curl/curl/issues/885
489 HOST is a command for a client to tell which host name to use, to offer FTP
490 servers named-based virtual hosting:
492 https://tools.ietf.org/html/rfc7151
494 4.2 Alter passive/active on failure and retry
496 When trying to connect passively to a server which only supports active
497 connections, libcurl returns CURLE_FTP_WEIRD_PASV_REPLY and closes the
498 connection. There could be a way to fallback to an active connection (and
499 vice versa). https://curl.haxx.se/bug/feature.cgi?id=1754793
501 4.3 Earlier bad letter detection
503 Make the detection of (bad) %0d and %0a codes in FTP URL parts earlier in the
504 process to avoid doing a resolve and connect in vain.
506 4.4 REST for large files
508 REST fix for servers not behaving well on >2GB requests. This should fail if
509 the server doesn't set the pointer to the requested index. The tricky
510 (impossible?) part is to figure out if the server did the right thing or not.
514 FTP ASCII transfers do not follow RFC959. They don't convert the data
517 4.6 GSSAPI via Windows SSPI
519 In addition to currently supporting the SASL GSSAPI mechanism (Kerberos V5)
520 via third-party GSS-API libraries, such as Heimdal or MIT Kerberos, also add
521 support for GSSAPI authentication via Windows SSPI.
523 4.7 STAT for LIST without data connection
525 Some FTP servers allow STAT for listing directories instead of using LIST,
526 and the response is then sent over the control connection instead of as the
527 otherwise usedw data connection: http://www.nsftools.com/tips/RawFTP.htm#STAT
529 This is not detailed in any FTP specification.
531 4.8 Option to ignore private IP addresses in PASV response
533 Some servers respond with and some other FTP client implementations can
534 ignore private (RFC 1918 style) IP addresses when received in PASV responses.
535 To consider for libcurl as well. See https://github.com/curl/curl/issues/1455
539 5.1 Better persistency for HTTP 1.0
541 "Better" support for persistent connections over HTTP 1.0
542 https://curl.haxx.se/bug/feature.cgi?id=1089001
544 5.2 support FF3 sqlite cookie files
546 Firefox 3 is changing from its former format to a a sqlite database instead.
547 We should consider how (lib)curl can/should support this.
548 https://curl.haxx.se/bug/feature.cgi?id=1871388
550 5.3 Rearrange request header order
552 Server implementors often make an effort to detect browser and to reject
553 clients it can detect to not match. One of the last details we cannot yet
554 control in libcurl's HTTP requests, which also can be exploited to detect
555 that libcurl is in fact used even when it tries to impersonate a browser, is
556 the order of the request headers. I propose that we introduce a new option in
557 which you give headers a value, and then when the HTTP request is built it
558 sorts the headers based on that number. We could then have internally created
559 headers use a default value so only headers that need to be moved have to be
564 Add the ability to specify the preferred authentication mechanism to use by
565 using ;auth=<mech> in the login part of the URL.
569 http://test:pass;auth=NTLM@example.com would be equivalent to specifying --user
570 test:pass;auth=NTLM or --user test:pass --ntlm from the command line.
572 Additionally this should be implemented for proxy base URLs as well.
574 5.6 Refuse "downgrade" redirects
576 See https://github.com/curl/curl/issues/226
578 Consider a way to tell curl to refuse to "downgrade" protocol with a redirect
579 and/or possibly a bit that refuses redirect to change protocol completely.
583 The standardization process of QUIC has been taken to the IETF and can be
584 followed on the [IETF QUIC Mailing
585 list](https://www.ietf.org/mailman/listinfo/quic). I'd like us to get on the
586 bandwagon. Ideally, this would be done with a separate library/project to
587 handle the binary/framing layer in a similar fashion to how HTTP/2 is
588 implemented. This, to allow other projects to benefit from the work and to
589 thus broaden the interest and chance of others to participate.
591 5.8 Leave secure cookies alone
593 Non-secure origins (HTTP sites) should not be allowed to set or modify
594 cookies with the 'secure' property:
596 https://tools.ietf.org/html/draft-ietf-httpbis-cookie-alone-01
603 Reading input (to send to the remote server) on stdin is a crappy solution for
604 library purposes. We need to invent a good way for the application to be able
605 to provide the data to send.
607 6.2 ditch telnet-specific select
609 Move the telnet support's network select() loop go away and merge the code
610 into the main transfer loop. Until this is done, the multi interface won't
613 6.3 feature negotiation debug data
615 Add telnet feature negotiation data to the debug callback as header data.
622 Add support for pipelining emails.
624 7.2 Enhanced capability support
626 Add the ability, for an application that uses libcurl, to obtain the list of
627 capabilities returned from the EHLO command.
629 7.3 Add CURLOPT_MAIL_CLIENT option
631 Rather than use the URL to specify the mail client string to present in the
632 HELO and EHLO commands, libcurl should support a new CURLOPT specifically for
633 specifying this data as the URL is non-standard and to be honest a bit of a
636 Please see the following thread for more information:
637 https://curl.haxx.se/mail/lib-2012-05/0178.html
644 Add support for pipelining commands.
646 8.2 Enhanced capability support
648 Add the ability, for an application that uses libcurl, to obtain the list of
649 capabilities returned from the CAPA command.
653 9.1 Enhanced capability support
655 Add the ability, for an application that uses libcurl, to obtain the list of
656 capabilities returned from the CAPABILITY command.
660 10.1 SASL based authentication mechanisms
662 Currently the LDAP module only supports ldap_simple_bind_s() in order to bind
663 to an LDAP server. However, this function sends username and password details
664 using the simple authentication mechanism (as clear text). However, it should
665 be possible to use ldap_bind_s() instead specifying the security context
666 information ourselves.
670 11.1 File listing support
672 Add support for listing the contents of a SMB share. The output should probably
673 be the same as/similar to FTP.
675 11.2 Honor file timestamps
677 The timestamp of the transferred file should reflect that of the original file.
681 Currently the SMB authentication uses NTLMv1.
683 11.4 Create remote directories
685 Support for creating remote directories when uploading a file to a directory
686 that doesn't exist on the server, just like --ftp-create-dirs.
692 There's no RFC for the protocol or an URI/URL format. An implementation
693 should most probably use an existing rsync library, such as librsync.
697 13.1 Disable specific versions
699 Provide an option that allows for disabling specific SSL versions, such as
700 SSLv2 https://curl.haxx.se/bug/feature.cgi?id=1767276
702 13.2 Provide mutex locking API
704 Provide a libcurl API for setting mutex callbacks in the underlying SSL
705 library, so that the same application code can use mutex-locking
706 independently of OpenSSL or GnutTLS being used.
708 13.3 Support in-memory certs/ca certs/keys
710 You can specify the private and public keys for SSH/SSL as file paths. Some
711 programs want to avoid using files and instead just pass them as in-memory
712 data blobs. There's probably a challenge to make this work across the
713 plethory of different TLS and SSH backends that curl suppports.
714 https://github.com/curl/curl/issues/2310
716 13.4 Cache/share OpenSSL contexts
718 "Look at SSL cafile - quick traces look to me like these are done on every
719 request as well, when they should only be necessary once per SSL context (or
720 once per handle)". The major improvement we can rather easily do is to make
721 sure we don't create and kill a new SSL "context" for every request, but
722 instead make one for every connection and re-use that SSL context in the same
723 style connections are re-used. It will make us use slightly more memory but
724 it will libcurl do less creations and deletions of SSL contexts.
726 Technically, the "caching" is probably best implemented by getting added to
727 the share interface so that easy handles who want to and can reuse the
728 context specify that by sharing with the right properties set.
730 https://github.com/curl/curl/issues/1110
732 13.5 Export session ids
734 Add an interface to libcurl that enables "session IDs" to get
735 exported/imported. Cris Bailiff said: "OpenSSL has functions which can
736 serialise the current SSL state to a buffer of your choice, and recover/reset
737 the state from such a buffer at a later date - this is used by mod_ssl for
738 apache to implement and SSL session ID cache".
740 13.6 Provide callback for cert verification
742 OpenSSL supports a callback for customised verification of the peer
743 certificate, but this doesn't seem to be exposed in the libcurl APIs. Could
744 it be? There's so much that could be done if it were!
746 13.7 improve configure --with-ssl
748 make the configure --with-ssl option first check for OpenSSL, then GnuTLS,
753 DNS-Based Authentication of Named Entities (DANE) is a way to provide SSL
754 keys and certs over DNS using DNSSEC as an alternative to the CA model.
755 https://www.rfc-editor.org/rfc/rfc6698.txt
757 An initial patch was posted by Suresh Krishnaswamy on March 7th 2013
758 (https://curl.haxx.se/mail/lib-2013-03/0075.html) but it was a too simple
759 approach. See Daniel's comments:
760 https://curl.haxx.se/mail/lib-2013-03/0103.html . libunbound may be the
761 correct library to base this development on.
763 Björn Stenberg wrote a separate initial take on DANE that was never
766 13.11 Support intermediate & root pinning for PINNEDPUBLICKEY
768 CURLOPT_PINNEDPUBLICKEY does not consider the hashes of intermediate & root
769 certificates when comparing the pinned keys. Therefore it is not compatible
770 with "HTTP Public Key Pinning" as there also intermediate and root certificates
771 can be pinned. This is very useful as it prevents webadmins from "locking
772 themself out of their servers".
774 Adding this feature would make curls pinning 100% compatible to HPKP and allow
775 more flexible pinning.
779 "HTTP Strict Transport Security" is TOFU (trust on first use), time-based
780 features indicated by a HTTP header send by the webserver. It is widely used
781 in browsers and it's purpose is to prevent insecure HTTP connections after
782 a previous HTTPS connection. It protects against SSLStripping attacks.
784 Doc: https://developer.mozilla.org/en-US/docs/Web/Security/HTTP_strict_transport_security
785 RFC 6797: https://tools.ietf.org/html/rfc6797
789 "HTTP Public Key Pinning" is TOFU (trust on first use), time-based
790 features indicated by a HTTP header send by the webserver. It's purpose is
791 to prevent Man-in-the-middle attacks by trusted CAs by allowing webadmins
792 to specify which CAs/certificates/public keys to trust when connection to
795 It can be build based on PINNEDPUBLICKEY.
797 Wikipedia: https://en.wikipedia.org/wiki/HTTP_Public_Key_Pinning
798 OWASP: https://www.owasp.org/index.php/Certificate_and_Public_Key_Pinning
799 Doc: https://developer.mozilla.org/de/docs/Web/Security/Public_Key_Pinning
800 RFC: https://tools.ietf.org/html/draft-ietf-websec-key-pinning-21
804 14.1 SSL engine stuff
806 Is this even possible?
808 14.2 check connection
810 Add a way to check if the connection seems to be alive, to correspond to the
811 SSL_peak() way we use with OpenSSL.
815 15.1 Add support for client certificate authentication
817 WinSSL/SChannel currently makes use of the OS-level system and user
818 certificate and private key stores. This does not allow the application
819 or the user to supply a custom client certificate using curl or libcurl.
821 Therefore support for the existing -E/--cert and --key options should be
822 implemented by supplying a custom certificate to the SChannel APIs, see:
823 - Getting a Certificate for Schannel
824 https://msdn.microsoft.com/en-us/library/windows/desktop/aa375447.aspx
826 15.2 Add support for custom server certificate validation
828 WinSSL/SChannel currently makes use of the OS-level system and user
829 certificate trust store. This does not allow the application or user to
830 customize the server certificate validation process using curl or libcurl.
832 Therefore support for the existing --cacert or --capath options should be
833 implemented by supplying a custom certificate to the SChannel APIs, see:
834 - Getting a Certificate for Schannel
835 https://msdn.microsoft.com/en-us/library/windows/desktop/aa375447.aspx
837 15.3 Add support for the --ciphers option
839 The cipher suites used by WinSSL/SChannel are configured on an OS-level
840 instead of an application-level. This does not allow the application or
841 the user to customize the configured cipher suites using curl or libcurl.
843 Therefore support for the existing --ciphers option should be implemented
844 by mapping the OpenSSL/GnuTLS cipher suites to the SChannel APIs, see
845 - Specifying Schannel Ciphers and Cipher Strengths
846 https://msdn.microsoft.com/en-us/library/windows/desktop/aa380161.aspx
850 16.1 Other authentication mechanisms
852 Add support for other authentication mechanisms such as OLP,
853 GSS-SPNEGO and others.
855 16.2 Add QOP support to GSSAPI authentication
857 Currently the GSSAPI authentication only supports the default QOP of auth
858 (Authentication), whilst Kerberos V5 supports both auth-int (Authentication
859 with integrity protection) and auth-conf (Authentication with integrity and
862 16.3 Support binary messages (i.e.: non-base64)
864 Mandatory to support LDAP SASL authentication.
871 SSH is a perfectly fine multiplexed protocols which would allow libcurl to do
872 multiple parallel transfers from the same host using the same connection,
873 much in the same spirit as HTTP/2 does. libcurl however does not take
874 advantage of that ability but will instead always create a new connection for
875 new transfers even if an existing connection already exists to the host.
877 To fix this, libcurl would have to detect an existing connection and "attach"
878 the new transfer to the existing one.
880 17.2 SFTP performance
882 libcurl's SFTP transfer performance is sub par and can be improved, mostly by
883 the approach mentioned in "1.6 Modified buffer size approach".
885 17.3 Support better than MD5 hostkey hash
887 libcurl offers the CURLOPT_SSH_HOST_PUBLIC_KEY_MD5 option for verifying the
888 server's key. MD5 is generally being deprecated so we should implement
889 support for stronger hashing algorithms. libssh2 itself is what provides this
890 underlying functionality and it supports at least SHA-1 as an alternative.
891 SHA-1 is also being deprecated these days so we should consider workign with
892 libssh2 to instead offer support for SHA-256 or similar.
894 17.4 Support CURLOPT_PREQUOTE
896 The two other QUOTE options are supported for SFTP, but this was left out for
899 18. Command line tool
903 "curl --sync http://example.com/feed[1-100].rss" or
904 "curl --sync http://example.net/{index,calendar,history}.html"
906 Downloads a range or set of URLs using the remote name, but only if the
907 remote file is newer than the local file. A Last-Modified HTTP date header
908 should also be used to set the mod date on the downloaded file.
912 Globbing support for -d and -F, as in 'curl -d "name=foo[0-9]" URL'.
913 This is easily scripted though.
915 18.3 prevent file overwriting
917 Add an option that prevents curl from overwriting existing local files. When
918 used, and there already is an existing file with the target file name
919 (either -O or -o), a number should be appended (and increased if already
920 existing). So that index.html becomes first index.html.1 and then
923 18.4 simultaneous parallel transfers
925 The client could be told to use maximum N simultaneous parallel transfers and
926 then just make sure that happens. It should of course not make more than one
927 connection to the same remote host. This would require the client to use the
928 multi interface. https://curl.haxx.se/bug/feature.cgi?id=1558595
930 Using the multi interface would also allow properly using parallel transfers
931 with HTTP/2 and supporting HTTP/2 server push from the command line.
933 18.5 UTF-8 filenames in Content-Disposition
935 RFC 6266 documents how UTF-8 names can be passed to a client in the
936 Content-Disposition header, and curl does not support this.
938 https://github.com/curl/curl/issues/1888
940 18.6 warning when setting an option
942 Display a warning when libcurl returns an error when setting an option.
943 This can be useful to tell when support for a particular feature hasn't been
944 compiled into the library.
946 18.7 warning if curl version is not in sync with libcurl version
948 This is usually a sign of a funny, weird or unexpected install situations
949 that aren't always quickly nor easily detected by users. curl and libcurl are
950 always released in sync and should use the same version numbers unless very
953 18.8 offer color-coded HTTP header output
955 By offering different color output on the header name and the header
956 contents, they could be made more readable and thus help users working on
959 18.9 Choose the name of file in braces for complex URLs
961 When using braces to download a list of URLs and you use complicated names
962 in the list of alternatives, it could be handy to allow curl to use other
965 Consider a way to offer that. Possibly like
966 {partURL1:name1,partURL2:name2,partURL3:name3} where the name following the
967 colon is the output name.
969 See https://github.com/curl/curl/issues/221
971 18.10 improve how curl works in a windows console window
973 If you pull the scrollbar when transferring with curl in a Windows console
974 window, the transfer is interrupted and can get disconnected. This can
975 probably be improved. See https://github.com/curl/curl/issues/322
977 18.11 -w output to stderr
979 -w is quite useful, but not to those of us who use curl without -o or -O
980 (such as for scripting through a higher level language). It would be nice to
981 have an option that is exactly like -w but sends it to stderr
982 instead. Proposed name: --write-stderr. See
983 https://github.com/curl/curl/issues/613
985 18.12 keep running, read instructions from pipe/socket
987 Provide an option that makes curl not exit after the last URL (or even work
988 without a given URL), and then make it read instructions passed on a pipe or
989 over a socket to make further instructions so that a second subsequent curl
990 invoke can talk to the still running instance and ask for transfers to get
991 done, and thus maintain its connection pool, DNS cache and more.
993 18.13 support metalink in http headers
995 Curl has support for downloading a metalink xml file, processing it, and then
996 downloading the target of the metalink. This is done via the --metalink option.
997 It would be nice if metalink also supported downloading via metalink
998 information that is stored in HTTP headers (RFC 6249). Theoretically this could
999 also be supported with the --metalink option.
1001 See https://tools.ietf.org/html/rfc6249
1003 See also https://lists.gnu.org/archive/html/bug-wget/2015-06/msg00034.html for
1004 an implematation of this in wget.
1006 18.14 --fail without --location should treat 3xx as a failure
1008 To allow a command line like this to detect a redirect and consider it a
1011 curl -v --fail -O https://example.com/curl-7.48.0.tar.gz
1013 ... --fail must treat 3xx responses as failures too. The least problematic
1014 way to implement this is probably to add that new logic in the command line
1015 tool only and not in the underlying CURLOPT_FAILONERROR logic.
1017 18.15 --retry should resume
1019 When --retry is used and curl actually retries transfer, it should use the
1020 already transferred data and do a resumed transfer for the rest (when
1021 possible) so that it doesn't have to transfer the same data again that was
1022 already transferred before the retry.
1024 See https://github.com/curl/curl/issues/1084
1026 18.16 send only part of --data
1028 When the user only wants to send a small piece of the data provided with
1029 --data or --data-binary, like when that data is a huge file, consider a way
1030 to specify that curl should only send a piece of that. One suggested syntax
1031 would be: "--data-binary @largefile.zip!1073741823-2147483647".
1033 See https://github.com/curl/curl/issues/1200
1035 18.17 consider file name from the redirected URL with -O ?
1037 When a user gives a URL and uses -O, and curl follows a redirect to a new
1038 URL, the file name is not extracted and used from the newly redirected-to URL
1039 even if the new URL may have a much more sensible file name.
1041 This is clearly documented and helps for security since there's no surprise
1042 to users which file name that might get overwritten. But maybe a new option
1043 could allow for this or maybe -J should imply such a treatment as well as -J
1044 already allows for the server to decide what file name to use so it already
1045 provides the "may overwrite any file" risk.
1047 This is extra tricky if the original URL has no file name part at all since
1048 then the current code path will error out with an error message, and we can't
1049 *know* already at that point if curl will be redirected to a URL that has a
1052 See https://github.com/curl/curl/issues/1241
1054 18.18 retry on network is unreachable
1056 The --retry option retries transfers on "transient failures". We later added
1057 --retry-connrefused to also retry for "connection refused" errors.
1059 Suggestions have been brought to also allow retry on "network is unreachable"
1060 errors and while totally reasonable, maybe we should consider a way to make
1061 this more configurable than to add a new option for every new error people
1064 https://github.com/curl/curl/issues/1603
1070 Consider extending 'roffit' to produce decent ASCII output, and use that
1071 instead of (g)nroff when building src/tool_hugehelp.c
1073 19.2 Enable PIE and RELRO by default
1075 Especially when having programs that execute curl via the command line, PIE
1076 renders the exploitation of memory corruption vulnerabilities a lot more
1077 difficult. This can be attributed to the additional information leaks being
1078 required to conduct a successful attack. RELRO, on the other hand, masks
1079 different binary sections like the GOT as read-only and thus kills a handful
1080 of techniques that come in handy when attackers are able to arbitrarily
1081 overwrite memory. A few tests showed that enabling these features had close
1082 to no impact, neither on the performance nor on the general functionality of
1090 Make our own version of stunnel for simple port forwarding to enable HTTPS
1091 and FTP-SSL tests without the stunnel dependency, and it could allow us to
1092 provide test tools built with either OpenSSL or GnuTLS
1094 20.2 nicer lacking perl message
1096 If perl wasn't found by the configure script, don't attempt to run the tests
1097 but explain something nice why it doesn't.
1099 20.3 more protocols supported
1101 Extend the test suite to include more protocols. The telnet could just do FTP
1102 or http operations (for which we have test servers).
1104 20.4 more platforms supported
1106 Make the test suite work on more platforms. OpenBSD and Mac OS. Remove
1107 fork()s and it should become even more portable.
1109 20.5 Add support for concurrent connections
1111 Tests 836, 882 and 938 were designed to verify that separate connections aren't
1112 used when using different login credentials in protocols that shouldn't re-use
1113 a connection under such circumstances.
1115 Unfortunately, ftpserver.pl doesn't appear to support multiple concurrent
1116 connections. The read while() loop seems to loop until it receives a disconnect
1117 from the client, where it then enters the waiting for connections loop. When
1118 the client opens a second connection to the server, the first connection hasn't
1119 been dropped (unless it has been forced - which we shouldn't do in these tests)
1120 and thus the wait for connections loop is never entered to receive the second
1123 20.6 Use the RFC6265 test suite
1125 A test suite made for HTTP cookies (RFC 6265) by Adam Barth is available at
1126 https://github.com/abarth/http-state/tree/master/tests
1128 It'd be really awesome if someone would write a script/setup that would run
1129 curl with that test suite and detect deviances. Ideally, that would even be
1130 incorporated into our regular test suite.
1133 21. Next SONAME bump
1135 21.1 http-style HEAD output for FTP
1137 #undef CURL_FTP_HTTPSTYLE_HEAD in lib/ftp.c to remove the HTTP-style headers
1138 from being output in NOBODY requests over FTP
1140 21.2 combine error codes
1142 Combine some of the error codes to remove duplicates. The original
1143 numbering should not be changed, and the old identifiers would be
1144 macroed to the new ones in an CURL_NO_OLDIES section to help with
1145 backward compatibility.
1147 Candidates for removal and their replacements:
1149 CURLE_FILE_COULDNT_READ_FILE => CURLE_REMOTE_FILE_NOT_FOUND
1151 CURLE_FTP_COULDNT_RETR_FILE => CURLE_REMOTE_FILE_NOT_FOUND
1153 CURLE_FTP_COULDNT_USE_REST => CURLE_RANGE_ERROR
1155 CURLE_FUNCTION_NOT_FOUND => CURLE_FAILED_INIT
1157 CURLE_LDAP_INVALID_URL => CURLE_URL_MALFORMAT
1159 CURLE_TFTP_NOSUCHUSER => CURLE_TFTP_ILLEGAL
1161 CURLE_TFTP_NOTFOUND => CURLE_REMOTE_FILE_NOT_FOUND
1163 CURLE_TFTP_PERM => CURLE_REMOTE_ACCESS_DENIED
1165 21.3 extend CURLOPT_SOCKOPTFUNCTION prototype
1167 The current prototype only provides 'purpose' that tells what the
1168 connection/socket is for, but not any protocol or similar. It makes it hard
1169 for applications to differentiate on TCP vs UDP and even HTTP vs FTP and
1172 22. Next major release
1174 22.1 cleanup return codes
1176 curl_easy_cleanup() returns void, but curl_multi_cleanup() returns a
1177 CURLMcode. These should be changed to be the same.
1179 22.2 remove obsolete defines
1181 remove obsolete defines from curl/curl.h
1185 make several functions use size_t instead of int in their APIs
1187 22.4 remove several functions
1189 remove the following functions from the public API:
1193 curl_mprintf (and variations)
1199 They will instead become curlx_ - alternatives. That makes the curl app
1200 still capable of using them, by building with them from source.
1202 These functions have no purpose anymore:
1206 curl_multi_socket_all
1208 22.5 remove CURLOPT_FAILONERROR
1210 Remove support for CURLOPT_FAILONERROR, it has gotten too kludgy and weird
1211 internally. Let the app judge success or not for itself.
1213 22.6 remove CURLOPT_DNS_USE_GLOBAL_CACHE
1215 Remove support for a global DNS cache. Anything global is silly, and we
1216 already offer the share interface for the same functionality but done
1219 22.7 remove progress meter from libcurl
1221 The internally provided progress meter output doesn't belong in the library.
1222 Basically no application wants it (apart from curl) but instead applications
1223 can and should do their own progress meters using the progress callback.
1225 The progress callback should then be bumped as well to get proper 64bit
1226 variable types passed to it instead of doubles so that big files work
1229 22.8 remove 'curl_httppost' from public
1231 curl_formadd() was made to fill in a public struct, but the fact that the
1232 struct is public is never really used by application for their own advantage
1233 but instead often restricts how the form functions can or can't be modified.
1235 Changing them to return a private handle will benefit the implementation and
1236 allow us much greater freedoms while still maintaining a solid API and ABI.