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 All bugs documented in the KNOWN_BUGS document are subject for fixing!
17 1.4 signal-based resolver timeouts
18 1.5 get rid of PATH_MAX
19 1.6 Modified buffer size approach
20 1.7 Detect when called from witin callbacks
21 1.8 Allow SSL (HTTPS) to proxy
23 2. libcurl - multi interface
25 2.2 Fix HTTP Pipelining for PUT
28 3.1 Update date and version in man paegs
32 4.2 Alter passive/active on failure and retry
33 4.3 Earlier bad letter detection
34 4.4 REST for large files
38 5.1 Better persistency for HTTP 1.0
39 5.2 support FF3 sqlite cookie files
40 5.3 Rearrange request header order
46 6.2 ditch telnet-specific select
47 6.3 feature negotiation debug data
48 6.4 send data in chunks
52 7.2 Enhanced capability support
56 8.2 Enhanced capability support
59 9.1 Enhanced capability support
62 10.1 SASL based authentication mechanisms
68 12.1 Disable specific versions
69 12.2 Provide mutex locking API
70 12.3 Evaluate SSL patches
71 12.4 Cache OpenSSL contexts
72 12.5 Export session ids
73 12.6 Provide callback for cert verification
74 12.7 improve configure --with-ssl
82 14.1 Other authentication mechanisms
87 15.3 prevent file overwriting
88 15.4 simultaneous parallel transfers
89 15.5 provide formpost headers
90 15.6 warning when setting an option
97 17.2 nicer lacking perl message
98 17.3 more protocols supported
99 17.4 more platforms supported
102 18.1 http-style HEAD output for ftp
103 18.2 combine error codes
104 18.3 extend CURLOPT_SOCKOPTFUNCTION prototype
106 19. Next major release
107 19.1 cleanup return codes
108 19.2 remove obsolete defines
110 19.4 remove several functions
111 19.5 remove CURLOPT_FAILONERROR
112 19.6 remove CURLOPT_DNS_USE_GLOBAL_CACHE
113 19.7 remove progress meter from libcurl
114 19.8 remove 'curl_httppost' from public
115 19.9 have form functions use CURL handle argument
116 19.10 Add CURLOPT_MAIL_CLIENT option
118 ==============================================================================
122 1.2 More data sharing
124 curl_share_* functions already exist and work, and they can be extended to
125 share more. For example, enable sharing of the ares channel and the
130 Use 'struct lifreq' and SIOCGLIFADDR instead of 'struct ifreq' and
131 SIOCGIFADDR on newer Solaris versions as they claim the latter is obsolete.
132 To support ipv6 interface addresses for network interfaces properly.
134 1.4 signal-based resolver timeouts
136 libcurl built without an asynchronous resolver library uses alarm() to time
137 out DNS lookups. When a timeout occurs, this causes libcurl to jump from the
138 signal handler back into the library with a sigsetjmp, which effectively
139 causes libcurl to continue running within the signal handler. This is
140 non-portable and could cause problems on some platforms. A discussion on the
141 problem is available at http://curl.haxx.se/mail/lib-2008-09/0197.html
143 Also, alarm() provides timeout resolution only to the nearest second. alarm
144 ought to be replaced by setitimer on systems that support it.
146 1.5 get rid of PATH_MAX
148 Having code use and rely on PATH_MAX is not nice:
149 http://insanecoding.blogspot.com/2007/11/pathmax-simply-isnt.html
151 Currently the SSH based code uses it a bit, but to remove PATH_MAX from there
152 we need libssh2 to properly tell us when we pass in a too small buffer and
153 its current API (as of libssh2 1.2.7) doesn't.
155 1.6 Modified buffer size approach
157 Current libcurl allocates a fixed 16K size buffer for download and an
158 additional 16K for upload. They are always unconditionally part of the easy
159 handle. If CRLF translations are requested, an additional 32K "scratch
160 buffer" is allocated. A total of 64K transfer buffers in the worst case.
162 First, while the handles are not actually in use these buffers could be freed
163 so that lingering handles just kept in queues or whatever waste less memory.
165 Secondly, SFTP is a protocol that needs to handle many ~30K blocks at once
166 since each need to be individually acked and therefore libssh2 must be
167 allowed to send (or receive) many separate ones in parallel to achieve high
168 transfer speeds. A current libcurl build with a 16K buffer makes that
169 impossible, but one with a 512K buffer will reach MUCH faster transfers. But
170 allocating 512K unconditionally for all buffers just in case they would like
171 to do fast SFTP transfers at some point is not a good solution either.
173 Dynamically allocate buffer size depending on protocol in use in combination
174 with freeing it after each individual transfer? Other suggestions?
176 1.7 Detect when called from witin callbacks
178 We should set a state variable before calling callbacks, so that we
179 subsequently can add code within libcurl that returns error if called within
180 callbacks for when that's not supported.
182 1.8 Allow SSL (HTTPS) to proxy
184 To prevent local users from snooping on your traffic to the proxy. Supported
186 http://www.chromium.org/developers/design-documents/secure-web-proxy
189 2. libcurl - multi interface
191 2.1 More non-blocking
193 Make sure we don't ever loop because of non-blocking sockets returning
194 EWOULDBLOCK or similar. Blocking cases include:
196 - Name resolves on non-windows unless c-ares is used
197 - NSS SSL connections
198 - HTTP proxy CONNECT operations
199 - SOCKS proxy handshakes
202 - The "DONE" operation (post transfer protocol-specific actions) for the
203 protocols SFTP, SMTP, FTP. Fixing Curl_done() for this is a worthy task.
205 2.2 Fix HTTP Pipelining for PUT
207 HTTP Pipelining can be a way to greatly enhance performance for multiple
208 serial requests and currently libcurl only supports that for HEAD and GET
209 requests but it should also be possible for PUT.
213 3.1 Update date and version in man pages
215 'maketgz' or another suitable script could update the .TH sections of the man
216 pages at release time to use the current date and curl/libcurl version
223 HOST is a suggested command in the works for a client to tell which host name
224 to use, to offer FTP servers named-based virtual hosting:
226 http://tools.ietf.org/html/draft-hethmon-mcmurray-ftp-hosts-11
228 4.2 Alter passive/active on failure and retry
230 When trying to connect passively to a server which only supports active
231 connections, libcurl returns CURLE_FTP_WEIRD_PASV_REPLY and closes the
232 connection. There could be a way to fallback to an active connection (and
233 vice versa). http://curl.haxx.se/bug/feature.cgi?id=1754793
235 4.3 Earlier bad letter detection
237 Make the detection of (bad) %0d and %0a codes in FTP url parts earlier in the
238 process to avoid doing a resolve and connect in vain.
240 4.4 REST for large files
242 REST fix for servers not behaving well on >2GB requests. This should fail if
243 the server doesn't set the pointer to the requested index. The tricky
244 (impossible?) part is to figure out if the server did the right thing or not.
248 FTP ASCII transfers do not follow RFC959. They don't convert the data
253 5.1 Better persistency for HTTP 1.0
255 "Better" support for persistent connections over HTTP 1.0
256 http://curl.haxx.se/bug/feature.cgi?id=1089001
258 5.2 support FF3 sqlite cookie files
260 Firefox 3 is changing from its former format to a a sqlite database instead.
261 We should consider how (lib)curl can/should support this.
262 http://curl.haxx.se/bug/feature.cgi?id=1871388
264 5.3 Rearrange request header order
266 Server implementors often make an effort to detect browser and to reject
267 clients it can detect to not match. One of the last details we cannot yet
268 control in libcurl's HTTP requests, which also can be exploited to detect
269 that libcurl is in fact used even when it tries to impersonate a browser, is
270 the order of the request headers. I propose that we introduce a new option in
271 which you give headers a value, and then when the HTTP request is built it
272 sorts the headers based on that number. We could then have internally created
273 headers use a default value so only headers that need to be moved have to be
278 Chrome and Firefox already support SPDY and lots of web services do. There's
279 a library for us to use for this (spdylay) that has a similar API and the
280 same author as nghttp2.
282 spdylay: https://github.com/tatsuhiro-t/spdylay
286 Add the ability to specify the preferred authentication mechanism to use by
287 using ;auth=<mech> in the login part of the URL.
291 http://test:pass;auth=NTLM@example.com would be equivalent to specifing --user
292 test:pass;auth=NTLM or --user test:pass --ntlm from the command line.
294 Additionally this should be implemented for proxy base URLs as well.
300 Reading input (to send to the remote server) on stdin is a crappy solution for
301 library purposes. We need to invent a good way for the application to be able
302 to provide the data to send.
304 6.2 ditch telnet-specific select
306 Move the telnet support's network select() loop go away and merge the code
307 into the main transfer loop. Until this is done, the multi interface won't
310 6.3 feature negotiation debug data
312 Add telnet feature negotiation data to the debug callback as header data.
314 6.4 send data in chunks
316 Currently, telnet sends data one byte at a time. This is fine for interactive
317 use, but inefficient for any other. Sent data should be sent in larger
324 Add support for pipelining emails.
326 7.2 Enhanced capability support
328 Add the ability, for an application that uses libcurl, to obtain the list of
329 capabilities returned from the EHLO command.
335 Add support for pipelining commands.
337 8.2 Enhanced capability support
339 Add the ability, for an application that uses libcurl, to obtain the list of
340 capabilities returned from the CAPA command.
344 9.1 Enhanced capability support
346 Add the ability, for an application that uses libcurl, to obtain the list of
347 capabilities returned from the CAPABILITY command.
351 10.1 SASL based authentication mechanisms
353 Currently the LDAP module only supports ldap_simple_bind_s() in order to bind
354 to an LDAP server. However, this function sends username and password details
355 using the simple authentication mechanism (as clear text). However, it should
356 be possible to use ldap_bind_s() instead specifing the security context
357 information ourselves.
363 There's no RFC for the protocol or an URI/URL format. An implementation
364 should most probably use an existing rsync library, such as librsync.
368 12.1 Disable specific versions
370 Provide an option that allows for disabling specific SSL versions, such as
371 SSLv2 http://curl.haxx.se/bug/feature.cgi?id=1767276
373 12.2 Provide mutex locking API
375 Provide a libcurl API for setting mutex callbacks in the underlying SSL
376 library, so that the same application code can use mutex-locking
377 independently of OpenSSL or GnutTLS being used.
379 12.3 Evaluate SSL patches
381 Evaluate/apply Gertjan van Wingerde's SSL patches:
382 http://curl.haxx.se/mail/lib-2004-03/0087.html
384 12.4 Cache OpenSSL contexts
386 "Look at SSL cafile - quick traces look to me like these are done on every
387 request as well, when they should only be necessary once per ssl context (or
388 once per handle)". The major improvement we can rather easily do is to make
389 sure we don't create and kill a new SSL "context" for every request, but
390 instead make one for every connection and re-use that SSL context in the same
391 style connections are re-used. It will make us use slightly more memory but
392 it will libcurl do less creations and deletions of SSL contexts.
394 12.5 Export session ids
396 Add an interface to libcurl that enables "session IDs" to get
397 exported/imported. Cris Bailiff said: "OpenSSL has functions which can
398 serialise the current SSL state to a buffer of your choice, and recover/reset
399 the state from such a buffer at a later date - this is used by mod_ssl for
400 apache to implement and SSL session ID cache".
402 12.6 Provide callback for cert verification
404 OpenSSL supports a callback for customised verification of the peer
405 certificate, but this doesn't seem to be exposed in the libcurl APIs. Could
406 it be? There's so much that could be done if it were!
408 12.7 improve configure --with-ssl
410 make the configure --with-ssl option first check for OpenSSL, then GnuTLS,
415 DNS-Based Authentication of Named Entities (DANE) is a way to provide SSL
416 keys and certs over DNS using DNSSEC as an alternative to the CA model.
417 http://www.rfc-editor.org/rfc/rfc6698.txt
419 An initial patch was posted by Suresh Krishnaswamy on March 7th 2013
420 (http://curl.haxx.se/mail/lib-2013-03/0075.html) but it was a too simple
421 approach. See Daniel's comments:
422 http://curl.haxx.se/mail/lib-2013-03/0103.html . libunbound may be the
423 correct library to base this development on.
427 13.1 SSL engine stuff
429 Is this even possible?
431 13.2 check connection
433 Add a way to check if the connection seems to be alive, to correspond to the
434 SSL_peak() way we use with OpenSSL.
438 14.1 Other authentication mechanisms
440 Add support for GSSAPI to SMTP, POP3 and IMAP.
446 "curl --sync http://example.com/feed[1-100].rss" or
447 "curl --sync http://example.net/{index,calendar,history}.html"
449 Downloads a range or set of URLs using the remote name, but only if the
450 remote file is newer than the local file. A Last-Modified HTTP date header
451 should also be used to set the mod date on the downloaded file.
455 Globbing support for -d and -F, as in 'curl -d "name=foo[0-9]" URL'.
456 This is easily scripted though.
458 15.3 prevent file overwriting
460 Add an option that prevents cURL from overwriting existing local files. When
461 used, and there already is an existing file with the target file name
462 (either -O or -o), a number should be appended (and increased if already
463 existing). So that index.html becomes first index.html.1 and then
466 15.4 simultaneous parallel transfers
468 The client could be told to use maximum N simultaneous parallel transfers and
469 then just make sure that happens. It should of course not make more than one
470 connection to the same remote host. This would require the client to use the
471 multi interface. http://curl.haxx.se/bug/feature.cgi?id=1558595
473 15.5 provide formpost headers
475 Extending the capabilities of the multipart formposting. How about leaving
476 the ';type=foo' syntax as it is and adding an extra tag (headers) which
477 works like this: curl -F "coolfiles=@fil1.txt;headers=@fil1.hdr" where
478 fil1.hdr contains extra headers like
480 Content-Type: text/plain; charset=KOI8-R"
481 Content-Transfer-Encoding: base64
482 X-User-Comment: Please don't use browser specific HTML code
484 which should overwrite the program reasonable defaults (plain/text,
487 15.6 warning when setting an option
489 Display a warning when libcurl returns an error when setting an option.
490 This can be useful to tell when support for a particular feature hasn't been
491 compiled into the library.
497 Consider extending 'roffit' to produce decent ASCII output, and use that
498 instead of (g)nroff when building src/tool_hugehelp.c
504 Make our own version of stunnel for simple port forwarding to enable HTTPS
505 and FTP-SSL tests without the stunnel dependency, and it could allow us to
506 provide test tools built with either OpenSSL or GnuTLS
508 17.2 nicer lacking perl message
510 If perl wasn't found by the configure script, don't attempt to run the tests
511 but explain something nice why it doesn't.
513 17.3 more protocols supported
515 Extend the test suite to include more protocols. The telnet could just do ftp
516 or http operations (for which we have test servers).
518 17.4 more platforms supported
520 Make the test suite work on more platforms. OpenBSD and Mac OS. Remove
521 fork()s and it should become even more portable.
525 18.1 http-style HEAD output for ftp
527 #undef CURL_FTP_HTTPSTYLE_HEAD in lib/ftp.c to remove the HTTP-style headers
528 from being output in NOBODY requests over ftp
530 18.2 combine error codes
532 Combine some of the error codes to remove duplicates. The original
533 numbering should not be changed, and the old identifiers would be
534 macroed to the new ones in an CURL_NO_OLDIES section to help with
535 backward compatibility.
537 Candidates for removal and their replacements:
539 CURLE_FILE_COULDNT_READ_FILE => CURLE_REMOTE_FILE_NOT_FOUND
541 CURLE_FTP_COULDNT_RETR_FILE => CURLE_REMOTE_FILE_NOT_FOUND
543 CURLE_FTP_COULDNT_USE_REST => CURLE_RANGE_ERROR
545 CURLE_FUNCTION_NOT_FOUND => CURLE_FAILED_INIT
547 CURLE_LDAP_INVALID_URL => CURLE_URL_MALFORMAT
549 CURLE_TFTP_NOSUCHUSER => CURLE_TFTP_ILLEGAL
551 CURLE_TFTP_NOTFOUND => CURLE_REMOTE_FILE_NOT_FOUND
553 CURLE_TFTP_PERM => CURLE_REMOTE_ACCESS_DENIED
555 18.3 extend CURLOPT_SOCKOPTFUNCTION prototype
557 The current prototype only provides 'purpose' that tells what the
558 connection/socket is for, but not any protocol or similar. It makes it hard
559 for applications to differentiate on TCP vs UDP and even HTTP vs FTP and
562 19. Next major release
564 19.1 cleanup return codes
566 curl_easy_cleanup() returns void, but curl_multi_cleanup() returns a
567 CURLMcode. These should be changed to be the same.
569 19.2 remove obsolete defines
571 remove obsolete defines from curl/curl.h
575 make several functions use size_t instead of int in their APIs
577 19.4 remove several functions
579 remove the following functions from the public API:
583 curl_mprintf (and variations)
589 They will instead become curlx_ - alternatives. That makes the curl app
590 still capable of using them, by building with them from source.
592 These functions have no purpose anymore:
596 curl_multi_socket_all
598 19.5 remove CURLOPT_FAILONERROR
600 Remove support for CURLOPT_FAILONERROR, it has gotten too kludgy and weird
601 internally. Let the app judge success or not for itself.
603 19.6 remove CURLOPT_DNS_USE_GLOBAL_CACHE
605 Remove support for a global DNS cache. Anything global is silly, and we
606 already offer the share interface for the same functionality but done
609 19.7 remove progress meter from libcurl
611 The internally provided progress meter output doesn't belong in the library.
612 Basically no application wants it (apart from curl) but instead applications
613 can and should do their own progress meters using the progress callback.
615 The progress callback should then be bumped as well to get proper 64bit
616 variable types passed to it instead of doubles so that big files work
619 19.8 remove 'curl_httppost' from public
621 curl_formadd() was made to fill in a public struct, but the fact that the
622 struct is public is never really used by application for their own advantage
623 but instead often restricts how the form functions can or can't be modified.
625 Changing them to return a private handle will benefit the implementation and
626 allow us much greater freedoms while still maintining a solid API and ABI.
628 19.9 have form functions use CURL handle argument
630 curl_formadd() and curl_formget() both currently have no CURL handle
631 argument, but both can use a callback that is set in the easy handle, and
632 thus curl_formget() with callback cannot function without first having
633 curl_easy_perform() (or similar) called - which is hard to grasp and a design
636 19.10 Add CURLOPT_MAIL_CLIENT option
638 Rather than use the URL to specify the mail client string to present in the
639 HELO and EHLO commands, libcurl should support a new CURLOPT specifically for
640 specifing this data as the URL is non-standard and to be honest a bit of a
643 Please see the following thread for more information:
644 http://curl.haxx.se/mail/lib-2012-05/0178.html