SPF_MAX_DNS_MECH is in multiple places
SPF_i_set_header_comment has a switch duplicated into spf_result.c
Test compile without pthreads and without res_ninit() - NetBSD 2

TODO items (somewhat in priority order)

* need to add better credits:
  * manu <at> netbsd.org (Emmanuel Dreyfus)
  * 

* SPF_dns_thread( lookup_layer, cache_layer )
  * read locking/write locking
  * spf layer function:  add_cache  (add to lowest cache or highest cache?)

* testing
  * domain name tests
    * sexy7/9.borg  (can this be a real problem?)
    * foo.com. (trailing dot in MX, PTR, etc.)
  * duplicate spf records
  * add checks to test for correct herrno handling
  * TXT w/ multiple strings
    * "v=spf1" "mx" "-all"
    * "v=spf1" " mx" " -all"
    * "v=spf1 " "mx " "-all "
    * "v=spf1 " "m" "x" " -all"
  * TXT w/ >512 bytes
  * DoS testing
  * "\"v=spf1 -all\""
  * -guess checks
  * break up test databases into smaller chunks
  * checks for nasty source routing and such for env-from
  * IPv6 tests
    * loopback
    * IPv4-mapped IPv6
    * a:/mx:
    * exists:
    * copy stuff from infradead.org
  * %{vr-}
  * envelope-from tests
    * multiple @
    * source routing
    * >63 char w/o dots in DNS lookups
  * ptr:foo.com should not match badfoo.com
  * add "extend example" tests from the spec


* spf_dns_zone:  option to read from file

* I don't think the stuff override/fallback sources are identified yet...

* IPv6 support
  * exists: IN6_IS_ADDR_V4MAPPED
  * rfc3596.txt says:
    * use AAAA records
    * use IP6.ARPA
  * rfc3513.txt  ::FFFF:d.d.d.d format
  * rfc3493.txt.gz discusses the ipv6 socket API

* do I implement best_guess correctly?

* add a string to SPF_compile_local_policy to change the "local
  policy" notation.  Use for DNSBL/RHSBL checks and such.  use a
  reason-text= modifier?

* spfquery -ip=192.0.2.200 -sender=50.spf1-test.mailzone.com -helo=50.spf1-test.mailzone.com
  * problem=Required option for mechanism missing near "" in "                                    -all";

* the get_spf and get_exp layers don't call the sublayers.

* investigate perlxs

* libspf API conversion layer  (what IS the libspf API??)

* document API

* improve debug output

* create release script

  * update Changelog

  * update version number in configure.ac
    AC_INIT(libspf_alt, 0.3.6, wayne@midwestcs.com)

  * update share library revisions in /lib/spf_alt/Makefile.am
    libspf_alt_la_LDFLAGS	= -version-info 0:0:0

  * write new announcement in doc

  * install new copy on backbone:/var/www/spf/libspf2/

  * save copy in /usr/local/src/spf/libspf2/

  * cd devel; make clean; make maintainer-clean; cd ..;
    tar czf libspf2.devel.0.3.tgz devel;
    scp libspf2.devel.0.3.tgz backbone:/home/wayne

* start using CVS

* implement a DNS timeout feature  (how?)

* implement a max time to eval feature

* SPF_dns_should_cache( mech )   if only str, %d, %r, %v used, return true

* code cleanup
  * many functions are way to long and need to be broken up
  * duplicate code
  * comments?  We don't need no stinkin' comments!
  * some files contain too many functions (e.g. spf_result.c)

* spfd
  * add logging
  * multi-threaded dns lookups  (how to cache?)
  * EOL string (currently uses \n, should it use \r\n?)
  * support for more commands?
  * spfquery.spfd
  * quote and escape characters

* SPF_verify()

     *   * version ok
     *   * lengths ok
     *   * counts ok
     *   * types ok
     *   * max_dns_mech
     *   * cidr data first and not both zero
     *   * ip4/ip6 cidr ranges
     *   * invalid chars in data_str and mod name
     *   * data on mechs that shouldn't have any
     *   * mechs that should have data, but don't


* fix all FIXME's

* convert int -> SPF_err_t
* convert int -> ns_type
* convert int -> SPF_dns_stat_t
* convert SPF_err_t, ns_type and SPF_dns_stat_t to enums

* document internals

* allow extentions to be turned on and off
  * redirect:
  * %{r}
  * %{c}
  * non-alpha chars in mech/mod
  * preserve case of localpart (new var?  use a delim?  overload URL encoding?)
  * exp-text= ?
  * support '@' as a delimeter?
  * document non-compliance
  * v=spf0, v=spf1-include or include-only=yes for use by includes
    and explanation?  

* need to make the system  more OS/compiler portable  (Windows, non-gcc)

* allow a sanatizer() function a-la M:S:Q?

* improve error messages from the compiler.
  * Sometimes the token is null, when it shouldn't be
  * Sometimes it is way too long.
    * 99txt.spf1-test.mailzone.com

* spf_dns_cid2spf

* add BEGIN_C_DECLS (autoconf book)  or __BEGIN_DECLS (in linux includes)

* use res_findzonecut to look for SPF records at the zone cut


* When the result of macro expansion is used in a domain name query, if
  the expanded domain name exceeds 255 characters (the maximum length
  of a domain name), the left side is truncated to fit, by removing
  successive subdomains until the total length falls below 255
  characters.

* it appears that the bits in the byte-compiled format are not laid
  out right.  I'm not sure that fields can cross byte boundaries

  bits/endian.h says: /* i386 is little-endian.  */

  /usr/include/arpa/nameser_compat.h
  ~/mail_rfcs/RFC/standard/rfc1035.txt.gz

  /usr/include/netinet/ip.h
 

* spfwhy.c utility to generate a web page
  * show all compile/lint errors
  * show results of evaluation
  * explain the SPF record

  * cgi script to use for an explanation
  * cgi script to parse and explain an SPF record (and give disgnostics)

* It would be better if all info about valid mechanisms was
  centeralized instead of being hard coded everywhere.

* spf_id2str needs to be broken up into smaller functions so that you
  can easily print a single mechanism, macro string, modifier, etc.

* make distcheck doesn't work

* make sure that macro-vars don't expand with trailing dots  (See spf-dev
  <7382FCA44E27D411BD4A00508BD68F950AE3102F@pigeon.tumbleweed.com>)