0N/AAPACHE 2.3 STATUS: -*-text-*-
381N/ALast modified at [$Date$]
0N/AThe current version of this file can be found at:
0N/ADocumentation status is maintained seperately and can be found at:
0N/AConsult the following STATUS files for information on related projects:
0N/APatches considered for backport are noted in their branches' STATUS:
0N/A 2.3.0 : in development
0N/AContributors looking for a mission:
0N/A * Just do an egrep on "TODO" or "XXX" in the source.
0N/A * Review the "PatchAvailable" bugs in the bug database:
0N/A After testing, you can append a comment saying "Reviewed and tested".
0N/A * Open bugs in the bug database.
0N/ACURRENT RELEASE NOTES:
0N/ARELEASE SHOWSTOPPERS:
0N/A * Handling of non-trailing / config by non-default handler is broken
0N/A jerenkrantz asks: Why should this block a release?
0N/A wsanchez agrees: this may be a change in behavior, but isn't
0N/A clearly wrong, and even if so, it doesn't seem like a
0N/A * the edge connection filter cannot be removed
0N/A jerenkrantz asks: Why should this block a release?
0N/A stas replies: because it requires a rewrite of the filters stack
0N/A implementation (you have suggested that) and once 2.2 is
0N/A released you can't do that anymore.
0N/A * If the parent process dies, should the remaining child processes
0N/A "gracefully" self-terminate. Or maybe we should make it a runtime
0N/A option, or have a concept of 2 parent processes (one being a
0N/A See: Message-ID: <3C58232C.FE91F19F@Golux.Com>
0N/A Self-destruct: Ken, Martin, Lars
0N/A Not self-destruct: BrianP, Ian, Cliff, BillS
0N/A Make it runtime configurable: Aaron, jim, Justin, wrowe, rederpj, nd
0N/A /* The below was a concept on *how* to handle the problem */
0N/A Have 2 parents: +1: jim
0N/A -1: Justin, wrowe, rederpj, nd
0N/A +0: Lars, Martin (while standing by, could it do
0N/A * Make the worker MPM the default MPM for threaded Unix boxes.
0N/A +1: Justin, Ian, Cliff, BillS, striker, wrowe, nd
0N/A +0: BrianP, Aaron (mutex contention is looking better with the
0N/A latest code, let's continue tuning and testing), rederpj, jim
0N/A pquerna: Do we want to change this for 2.2?
0N/ARELEASE NON-SHOWSTOPPERS BUT WOULD BE REAL NICE TO WRAP THESE UP:
0N/A * Patches submitted to the bug database:
0N/A * Filter stacks and subrequests, redirects and fast redirects.
0N/A There's at least one PR that suffers from the current unclean behaviour
0N/A (which lets the server send garbage): PR 17629
0N/A nd says: Every subrequest should get its own filter stack with the
0N/A subreq_core filter as bottom-most. That filter does two things:
0N/A - swallow EOS buckets
0N/A - redirect the data stream to the upper request's (rr->main)
0N/A filter chain directly after the subrequest's starting
858N/A Once we have a clean solution, we can try to optimize
0N/A it, so that the server won't be slow down too much.
0N/A * RFC 2616 violations.
0N/A Open PRs: 15852, 15859, 15861, 15864, 15865, 15866, 15868, 15869,
0N/A 15870, 16120, 16125, 16126, 16133, 16135, 16136, 16137,
0N/A 16138, 16139, 16140, 16142, 16518, 16520, 16521,
0N/A jerenkrantz says: need to decide how many we need to backport
and/or 0N/A if these rise to showstopper status.
0N/A wrowe suggests: it would be nice to see "MUST"
v.s. "SHOULD"
v.s. "MAY"
0N/A out of this list, without reviewing them individually.
0N/A * There is a bug in how we sort some hooks, at least the pre-config
0N/A hook. The first time we call the hooks, they are in the correct
0N/A order, but the second time, we don't sort them correctly. Currently,
0N/A back when this is fixed.
0N/A OtherBill offers that this is a SERIOUS problem. We do not sort
0N/A correctly by the ordering arguments passed to the register hook
0N/A functions. This was proven when I reordered the open_logs hook
0N/A to attempt to open the error logs prior to the access logs. Possibly
0N/A the entire sorting code needs to be refactored.
0N/A * pipes deadlock on all platforms with limited pipe buffers (
e.g. both
0N/A Linux and Win32, as opposed to only Win32 on 1.3). The right solution
0N/A is either GStein's proposal for a "CGI Brigade", or OtherBill's proposal
0N/A for "Poll Buckets" for "Polling Filter Chains". Or maybe both :-)
0N/A * All handlers should always send content down even if r->header_only
0N/A is set. If not, it means that the HEAD requests don't generate the
0N/A same headers as a GET which is wrong.
0N/A * exec cmd and suexec arg-passing enhancements
0N/A Status: Patches proposed
0N/A Message-ID: <20020526041748.A29148@prodigy.Redbrick.DCU.IE>
0N/A * The 2.0.36 worker MPM graceless shutdown changes work but are
0N/A a bit clunky on some platforms; eg, on Linux, the loop to
0N/A join each worker thread seems to hang, and the parent ends up
0N/A killing off the child with SIGKILL. But at least it shuts down.
0N/A * --enable-mods-shared="foo1 foo2" is busted on Darwin. Pier
0N/A posted a patch (Message-ID: <B8DBBE8D.575A%pier@betaversion.org>).
0N/A * We do not properly substitute the prefix-variables in the configuration
0N/A scripts or generated-configs. (
i.e. if sysconfdir is etc,
0N/A * If any request gets through ap_process_request_internal() and is
0N/A scheduled to be served by the core handler, without a flag that this
0N/A end of the ap_process_request_internal() processing so sub_req-esters
0N/A know this request cannot be run. This provides authors of older
0N/A modules better compatibility, while still improving the security and
0N/A Status: still need to decide where this goes, OtherBill comments...
0N/A Message-ID: <065701c14526$495203b0$96c0b0d0@roweclan.net>
0N/A [Deleted comments regarding the ap_run_handler phase, as irrelevant
0N/A as BillS points out that "common case will be caught in
0N/A and the issue is detecting this -before- we try to run the req.]
0N/A gregames says: can this happen somehow without a broken module
0N/A being involved? If not, why waste cycles trying to defend against
0N/A potential broken modules? It seems futile.
0N/A wrowe counters: no, it shouldn't happen unless the module is broken.
0N/A But the right answer is to fail the request up-front in
dir/file 0N/A walk if the path was entirely invalid; and we can't do that either
0N/A UNTIL 2.1 or we break modules that haven't hooked map_to_storage.
0N/A * With AP_MODE_EXHAUSTIVE in the core, it is finally clear to me
0N/A how the Perchild MPM should be re-written. It hasn't worked
0N/A correctly since filters were added because it wasn't possible to
0N/A get the content that had already been written and the socket at
0N/A the same time. This mode lets us do that, so the MPM can be
0N/A * Can a static httpd be built reliably?
0N/A Message-ID: <20020207142751.T31582@clove.org>
0N/A * Usage of APR_BRIGADE_NORMALIZE in core_input_filter should be
0N/A removed if possible.
0N/A Message-ID: <Pine.LNX.4.33.0201202232430.318-100000@deepthought.cs.virginia.edu>
0N/A Jeff wonders if we still care about this. It is no longer an
0N/A API issue but simply an extra trip through the brigade.
0N/A * Get perchild to work on platforms other than Linux. This
0N/A descriptors between vhost child groups. An API was proposed
0N/A Message-ID: <20020111115006.K1529@clove.org>
0N/A * Try to get libtool inter-library dependency code working on AIX.
0N/A Message-ID: <cm3n10lx555.fsf@rdu163-40-092.nc.rr.com>
0N/A Justin says: If we get it working on AIX, we can enable this
257N/A on all platforms and clean up our build system
257N/A Jeff says: I thought I tested a patch for you sometime in
257N/A January that you were going to commit within a few
257N/A * Handling of %2f in URIs. Currently both 1.3 and 2.0
257N/A completely disallow %2f in the request URI path (see
257N/A ap_unescape_url() in
util.c). It's permitted and passed
257N/A through in the query string, however. Roy says the
257N/A original reason for disallowing it, from five years ago,
257N/A was to protect CGI scripts that applied PATH_INFO to
257N/A a filesystem location and which might be tricked by
257N/A ..%2f..%2f(...). We *should* allow path-info of the
257N/A Since we've revamped a lot of our processing of path
257N/A segments, it would be nice to allow this, or at least
257N/A allow it conditionally with a directive.
257N/A OtherBill adds that %2f as the SECOND character of a multibyte
257N/A sequence causes the request to fail! This happens notably in
0N/A * FreeBSD, threads, and worker MPM. All seems to work fine
0N/A if you only have one worker process with many threads. Add
0N/A a second worker process and the accept lock seems to be
0N/A lost. This might be an APR issue with how it deals with
0N/A the child_init hook (
i.e. the fcntl lock needs to be resynced).
0N/A More examination and analysis is required.
0N/A Status: Works with FreeBSD 5.3. Does not work in previous versions.
0N/A This has also been reported on Cygwin.
0N/A * There is increasing demand from module writers for an API
0N/A that will allow them to control the server � la apachectl.
0N/A Reasons include sole-function servers that need to die if
0N/A an external dependency (
e.g., a database) fails, et cetera.
0N/A Perhaps something in the (ever more abused) scoreboard?
0N/A On the other hand, we already have a pipe that goes between parent
0N/A and child for graceful shutdown events, along with an API that
0N/A can be used to send a message down that pipe. In threaded MPMs,
0N/A it is easy enough to make that one pipe be used for graceful
0N/A and graceless events, and it is also easy to open that pipe
0N/A to both parent and child for writing. Then we just need to
0N/A figure out how to do graceless on non-threaded MPMs.
0N/A * Allow the DocumentRoot directive within <Location > scopes? This
0N/A and in-your-face.) DocumentRoot unset would be accepted [and would
0N/A not permit content to be served, only virtual resources such as
0N/A server-info or server-status.
0N/A This proposed change would _not_ depricate Alias.
0N/A striker: See the thread starting with Message-ID:
0N/A JLEGKKNELMHCJPNMOKHOGEEJFBAA.striker@apache.org.
0N/A * Win32: Rotatelogs sometimes is not terminated when Apache
0N/A goes down hard. FirstBill was looking at possibly tracking the
0N/A child's-child processes in the parent process.
0N/A stoddard: Shared scoreboard might offer a good way for the parent
0N/A to keep track of 'other child' processes and whack them if the child
0N/A Other thoughts on walking the process chain using the NT kernel
257N/A have also been proposed on APR.
257N/A * Eliminate unnecessary creation of pipes in mod_cgid
257N/A * Platforms that do not support fork (primarily Win32 and AS/400)
257N/A Architect start-up code that avoids initializing all the modules
257N/A in the parent process on platforms that do not support fork.
257N/A * There are still a number of places in the code where we are
257N/A losing error status (
i.e. throwing away the error returned by a
257N/A system call and replacing it with a generic error code)
1248N/A * Mass vhosting version of suEXEC.
1248N/A the dbmmanage employs the first-matched dbm format. This is not
1248N/A necessarily the library that Apache was built with. Aught to
1248N/A rewrite dbmmanage upon installation to bin/ with the proper library
1248N/A for predictable mod_auth_dbm administration.
257N/A Questions; htdbm exists, time to kill dbmmanage, or does it remain
257N/A useful as a perl dbm management example? If we keep it,
257N/A do we address the issue above?
257N/A Some additional items remaining:
257N/A - case_preserved_filename stuff
257N/A (use the new canonical name stuff?)
257N/A - find a new home for ap_text(_header)
257N/A - is it possible to remove the DAV: namespace stuff from util_xml?
257N/A * ap_core_translate() and its use by mod_mmap_static and mod_file_cache
257N/A are a bit wonky. The function should probably be exposed as a utility
257N/A function (such as ap_translate_url2fs() or ap_validate_fs_url() or
257N/A something). Another approach would be a new hook phase after
257N/A "translate" which would allow the module to munge what the
257N/A translation has decided to do.
257N/A Status: Greg +1 (volunteers)
257N/A calls ap_fixup_virutal_hosts(), ap_fini_vhost_config(), and
257N/A ap_sort_hooks() [to reduce the logic in main()]
257N/A * read the config tree just once, and process N times (as necessary)
257N/A * (possibly) use UUIDs in mod_unique_id
and/or mod_usertrack
0N/A * (possibly) port the bug fix for PR 6942 (segv when LoadModule is put
0N/A into a VirtualHost container) to 2.0.
0N/A * callers of ap_run_create_request() should check the return value
0N/A for failure (Doug volunteers)
0N/A * Fix the worker MPM to use POD to kill child processes instead
0N/A of ap_os_killpg, regardless of how they should die.
0N/A * Scoreboard structures could be changed in the future such that
0N/A proper alignment is not maintained, leading to segfaults on
0N/A some systems. Cliff posted a patch to deal with this issue but
0N/A later recanted. See this message to dev@apr.apache.org:
0N/A * APXS either needs to be fixed completely for use when apr is out of tree,
0N/A or it should drop query mode altogether, and we just grow an
0N/A httpd-config or similar arrangement.
0N/A To quote a discussion in STATUS earlier:
257N/A thommay: this doesn't fix all the problems with apxs and out of
257N/A query cases; but I'm beginning to think that in these cases
257N/A the app should be querying ap{r,u}-config directly
257N/A gstein: agreed. apxs should deprecate the -q flag
257N/A pquerna: I vote for a httpd-config, and to deprecate the -q flag.
257N/A minfrin: +1 for httpd-config, and to deprecate -q.
257N/ATODO ISSUES REMAINING IN MOD_SSL:
257N/A * In order to use a DSO version of mod_ssl we have to link with
257N/A -lssl and -lcrypto. A workaround is in place right now where the
257N/A entire EXTRA_LIBS macro is being appended to the objects list, but
257N/A this is a hack. We should either revamp the APACHE_CHECK_SSL_TOOLKIT
257N/A autoconf function or come up with some other autoconf checks to
257N/A search for libssl and libcrypto and properly add them to mod_ssl's
257N/A * SSL renegotiations in combination with POST request
257N/A * Port or dispose all code inside #if 0...#endif blocks that remain
257N/A from the porting effort.
257N/A * Do we need SSL_set_read_ahead()?
0N/A * the ssl_expr api is NOT THREAD SAFE. race conditions exist:
0N/A -in ssl_expr_comp() if SSLRequire is used in .htaccess
0N/A (ssl_expr_info is global)
0N/A -is ssl_expr_eval() if there is an error
0N/A (ssl_expr_error is global)
0N/A * SSLRequire directive (parsing of) leaks memory
0N/A * Diffie-Hellman-Parameters for temporary keys are hardcoded in
0N/A "it is suggested that keys be changed daily or every 500
0N/A transactions, and more often if possible."
0N/A * ssl_var_lookup could be rewritten to be MUCH faster
0N/A * CRL callback should be pluggable
0N/A * session cache store should be pluggable
0N/A * init functions should return status code rather than ssl_die()
0N/A to also decrypt proxy keys
0N/A * the shmcb code should just align its memory segment rather than
0N/A jumping through all the "safe" memcpy and memset hoops
0N/A * mod_proxy: Ability to run SSL over proxy gateway connections,
0N/A encrypting (or reencrypting) at the proxy.
0N/A * mod_cache: Handle ESI tags.
0N/A * mod_cache: Resolve issue of how to cache page fragements (or perhaps
0N/A will cache #include 'virtual' requests (but not #include 'file'
0N/A requests). This was accomplished by making CACHE_IN a
0N/A CONTENT_SET-1 filter to force it to run before the SUBREQ_CORE
0N/A filter. But now responses cannot be cached that include the
0N/A effects of having been run through CONTENT_SET filters
0N/A (mod_deflate, mod_expires, etc). We could rerun all the
0N/A CONTENT_SET filters on the cached response, but this will not
0N/A work in all cases. For example, mod_expires relies on installing
0N/A the EXPIRATION filter during fixups. Contents served out of
0N/A mod_cache (out of the quick_handler) bypass -all- the request
0N/A line server hooks (Ryan really hated this. It is great for
0N/A performance, but bad because of the complications listed above).
0N/A * mod_mem_cache: Consider adding a RevalidateTimeout directive to
0N/A specify time at which local cached content is to be revalidated
0N/A (ie, underlying file stat'ed to see if it has changed).
jerenkrantz says: Too slow. Get regexs away from speedy caches by
default. Introduce a new CacheEnableRegex if you want.
status (num of entries, cache object properties, etc.).
mod_status could be extended to query optional hooks defined
by modules for the purpose of reporting module status.
mod_cache (et. al.) could define optional hooks that are called
to collect status. Status should be queryable by
jerenkrantz says: Yawn. Who cares.
* MaxRequestsPerChild measures connections, not requests.
Until someone has a better way, we'll probably just rename it
"MaxConnectionsPerChild".
* Regex containers don't work in an intutive way
Status: No one has come up with an efficient way to fix this
behavior. Dean has suggested getting rid of regex containers
OtherBill suggests: We at least seem to agree on eliminating
the <Container ~ foo> forms, and using only
<ContainerMatch foo> semantics.
needed. Apache 1.3 just never stashed "multipart" into
r->content_type. We should probably follow suit since the
byterange stuff doesn't want the rest of the code to see the
multipart content-type; the other code should still think it is
dealing with the <orig_ct> stuff.
Status: Greg volunteers to investigate (esp. since he was most
likely the one to break it :-)
Experimental modules should eventually be be promoted to fully supported
status or removed from the repository entirely (ie, the
'experiment' failed). This section tracks what needs to happen to
get the modules promoted to fully supported status.