Note that some of these are almost certainly better off in seperate libraries.
Some are pretty high priority for the next stable version.
 I need a better priority ... but so far I've mainly been concerned with what
was needed for vstr-1.0.0

 Maybe During 1.0.x
 ------------------

. ____ make vstr_fmt.c faster ____

. vstr_spn_cstr_buf_fwd etc.

. test len=0 on _sc_*()

. read/write entire file with read_file/write_file

. glibc printf tests ... enable all three forms of fmt for coverage?

. examples/ex_relay

. examples/chatserv

. examples/ex_dnsd

. examples/conf

. Allow '-' filename, for stdin

. restrict  (speed)

. autodetect linker-script options (build)

. AM_CFLAGS (build)

. timeval custom formatter (feature)

. rfc2732 for ipv4/ipv6 addresses and port

. Some way to say use a _BUF if we are subbing in a _BUF and use a _PTR/_REF
  otherwise. Or a way to say how much data can I add into this _BUF node
 (Ie. currently, is it a _BUF and does it have free space). (feature/speed)

. malloc_bad might be able to be set when all of the functions have worked
  (Ie. position cache allocation used to fail) (testing).

. comparison.html ... portability plus ...
  can you copy and paste the code ... or need a shared lib. (docs)

. vstr_mov() doesn't move cache information. (speed)

. Get rid of inet_ntop() ? (portability)

. wcrtomb() doesn't exist on FreeBSD ... relies on LC_CTYPE (portability)

. del() inline kills cache too easily ... make cb's extern inline and call
   directly. (speed)

 1.2.x
 -----

. utf-8 character class, utf-8 searching etc.

 Misc
 ----

. Vstr_ref->ptr non const. but possibly never altered, kind of needed for free()
  calls *sigh*

# This is all one feature, I think... vstr_ref* API needs a new external home.
. Allocator - seperate library
. _SEF type ... like a _REF but you can do substitution on it when ref == 1,
  splitting the node changes behaviour here so probably not worth it *sigh*.
  Could do a copy when you move it to a different Vstr string, but that doesn't
 really do much more then _BUF nodes. Although you could convert _BUF nodes to
 it on a copy, which might be worth it.
  Could put a double indirect _ref in there ... with the _ref for the Vstr
 string and the second for all strings ... then allow subs if the second == 1.
   Will do "strange" things on mmap() read/write areas on _conv_ functions
  as it's not obvious what it should/could do on expansion.
. Vstr_lref which has a length? So we can implement...
   vstr_lref_*_malloc()
   vstr_lref_*_realloc()
  ...which will do the right thing if ref->ref == 1 ? (feature)
  This also implies some memory pool like semantics? (feature)
  Memory pools can kind of be done using a Vstr and adding the _REF nodes
 (mmap is hard ... as is sharing across strings).

. VSTR_FLAG_PARSE_NUM_LOC_SEP

. Add vstr_parse_double()/long_double()

. vstr_conv_canon_path() Ie. /foo/bar/../abcd -=> /foo/abcd

. http://burtleburtle.net/bob/hash/doobs.html .. vstr_x_hash ?
  vstr_x_hash_md5 vstr_x_hash_sha1

. vstr_x_pcre_*

. _Bool ?

. {S...:%s} for chopped strings to replace end with "...".

. _conv_tr_buf_buf() ? ...
   Ie. vstr_conv_tr_buf_buf(s1, 1, s1->len, "\t\r", 2, " ?", 2)

. _conv_sub_buf_buf() ? ...
   Ie. vstr_conv_sub_buf_buf(s1, 1, s1->len, "\t", 1, "    ", 4)

. Fast inline parsing of ints when using certain flags.

. Add --with-libc=diet,host,klibc -- put into test framework.

. vstr_add_fmt() internal cleanup to use struct blah {Vstr_fmt ... }

. ASN.1 (International Standard 8824)

. XDR ... rpc

. More locale related stuff, incl. vstr_loc_ functions for cmp/srch/etc.

. vstr_conv_encode_qp (quoted printable rfc2045.txt)
. vstr_conv_encode_b64 (base64 rfc2045.txt)