psyclpc/TODO

FROM LDMUD
- ldmud has made quite some progress, although not exactly in the areas
  that we are active in. syncing with ldmud is a good idea.

BUGS
- erq doesn't conform to fortify code safety standards. this is ugly and
  should be fixed but it is okay to just turn off fortify because erq is
  fed exclusively with sanitized data from psyced, so buffer overruns are
  possible but only theoretical.
- configure should warn more vehemently when libidn is missing
- libpsyc isnt recognized even if properly installed
- x86_64 seems to require -ldl explicitly at the end of libs
- should autoconf to sysmalloc also for osol (OpenSolaris)
- #define USE_EXPAT und JSON werden trotzdem gesetzt in config.h
  auch wenn configure keine libs gefunden hat (egal, wir verwenden beide nicht)
- Because of some funny bug the driver will probably not bind properly
  to _basic_host_IP as it also needs _basic_host_name and _basic_host_domain
  to be provided. Hopefully this shortcoming will soon be history.
- what's the use of a traceback on stdout if the actual error goes to stderr?
    Error loading place: *Error in loading object: 'place/testotest'.
    Warning: place "/place/testotest" did not return ME on load()
    2009.04.21 12:35:19 ERROR caught. Backtrace:
    '       parse' in '      net/irc/user.c' (' net/irc/user#User') line 32
    [...]
    '       CATCH' in                        ('drivers/ldmud/library/library')
- libpsyc detection fails on BSD, hotfix by adding #define HAS_PSYC manually
- BSD needs -liconv in generated makefile

WINDOWS SUPPORT
+ with the new __NO_SRV__ detection psyclpc compiles under cygwin but it
  doesn't provide SRV which makes it a suboptimal XMPP router.. just in
  case you want that.
      fippo suggests to put http://delta.affinix.com/jdns/ into psyclpc
  http://www.chiark.greenend.org.uk/~ian/adns/ requires a locally running
  nameserver.. and i don't think it even runs on cygwin.. other ideas?

COMPLAINTS
-* QA Notice: Package has poor programming practices which may compile
 *            fine but exhibit random runtime failures.
 * interpret.c:7224: warning: dereferencing type-punned pointer will break strict-aliasing rules

MISC IDEAS
- make more MUD features optional: more d_flag stuff?
  ? USE_RESETS time_of_ref USE_CLEANUP load_id USE_WIZLIST USE_EXPAT USE_EVALUATION_COST
  - disabling with_input_escape does not disable the input_escape handling source code
+ teach ldmud to exec() so that, together with --args, we can run it as a
  regular daemon without kludgy shellscripts. zesstra suggests we could have
  a C wrapper process instead of a shell script. and what about init, can't
  we monitor and restart the driver from init?
	(http://ldmud.eu/mantis/view.php?id=55)
  + provide a query_shutdown_progress() efun to inform about progress
    and the intention to either shutdown or restart. maybe even a reason?
+ optimize f_lower_case by adding an 'already lowercase' flag to the
  string data structure
+ ldap package erproben und fertigmachen
+ put errno() back in
+ psyctext() in C ... see psyced/CHANGESTODO for details
? bugless says there is an ioctl() to obtain a descriptor's owner uid
  thus avoid the authlocal /proc hack
? have a seperate package for the LPC documentation and update it a bit
  since ldmud?
+ fip thinks we should have a new operator with arguments rather than 
  clone_object and subsequent -> calls. pike doesn't even have the "new"
  because unlike C++ and LPC it forbids static use of classes (aka "blueprints")
+ fippo suggests to look into http://www.hpl.hp.com/personal/Hans_Boehm/gc/leak.html
+ DMSG_FLUSH for incremental debug messages  :)
- for pike compatibility allow string[-1] just like string[<1]
+ would be nice if sprintf %O would indent data structure according to depth.
  currently outputting a parsed xml node is quite confusing

ENCRYPTION (see http://about.psyc.eu/Talk:Encryption)
? link libpurple to this for protocol supports and OTR hahahahaha
? or link libotr only?
? or is OTR too expensive anyway and we should try SS-TLS?

USE_RESTORED_OBJECTS
 + object names need to be able to be identical to uniforms
   then it doesn't matter if they came back from persistence as strings
   as sendmsg() can handle both objects and uniforms - we just need to
   deal with mappings where both the object and the uniform could be the key..
    (members & routes of places)

NETWORKING
- add_message("%s", "\n*** Text lost in transmission ***\n");
  should be passed to master instead.
+ _length: <eL> somebody implement a new input_to(#'get_data, length).
? replace ERQ with c-ares http://c-ares.haxx.se/ for async dns lookups
? MAYBE the whole networking should be replaced by libevent!??
+ support epoll() / kqueue (or just libevent?)

? allow port definition from lpc:master.c or even dynamic listen()
  so we can add port config to the web configurator
..egal, machen wir jetzt über psyconf

CHARSET
-?  keeping track of charset everywhere to avoid any unnecessary conversion
    would lead to so much overhead, that occasional unnecessary conversion is
    the cheaper option. that's why we should convert in the driver close to
    the network operations:
comm.c:3055:    if(ip->tls_inited)
                    l = tls_read(ip, ip->text + ip->text_end, (size_t)l);
                else if (!(ip->tls_want))
                    l = socket_read(ip->socket, ip->text + ip->text_end, ...
+ since network read operations are much more efficient than input_to-lines
  iconv charset conversions should happen right after socket/tls_read()
  into a pre-buffer, then copy-converted into the interactive's buffer
+ outgoing transmissions need to find the same treatment, before
  socket_write() and also before optional compression.
+ socket charset is stored in interactive_s and given by an efun
+ charset is compared to SYSTEM_CHARSET which is given by config.h/configure
- http://ldmud.eu/mantis/view.php?id=117 ... try //TRANSLIT again?