Bug List
--------

--- Fixed on Build 001 ---

X Answers with Uppercase.
  Do not add blanks in front/back of tag evaluation results.
X <br/> ignored (fixed thru patch on XML Parser object), XML.Normalize->FALSE.
X <set_topic></set_topic> didn't work correctly (actually all <set_*.></set_*>
X Added wilcard (*) on <that></that>
X BOTSAID must store by sentence, will use '.' as a sentence separator and
  strip out <cr>,<lf>
X Bug fix in the <condition></condition> format (bad interpretation of AIML).
X Control a single instance of the program running at any time.


--- Fixed on Build 002 ---

X <that></that> must be able to refer to index <> "0".
  Format <that index="1">xxxx</that> (Pattern Side).

--- Fixed on Build 003 ---

X Bug on XML Object Parser, leading and trailing blanks were stripped off
  by normalization; that produced patterns and templates to be ruined
  (i.e. hello <get_name/> --> hello{name}). Special check added to avoid
  that problem (workaround of XML object behaviour).
X <li> without value attribute within <condition> means "if all other failed",
  first successful <li> interrupts evaluation, a <li> without a value attr
  is considered as "always match".

--- Fixed on Build 004 ---

X On <learn> if no FILE attribute was indicated it defaults to the NAME
  attribute (topic) overriding the posibility to learn a pattern/template
  (per-lola.aiml <pattern>*</pattern> structure).

X Bug on the pattern matching algorithm of <that></that>.

--- Fixed on Build 005 ---

X Adapt Web Interface to new criteria to store BotSaid/ClientSaid, an
  internal structure to the web responder stores the complete answer
  from the bot as it's provided by the engine without any sentence
  segmentation.

--- Fixed on Build 006 ---

X Investigate how to guarantee a space is introduced after a sentence.

X Web interface history is broken on multi-sentence.


--- Fixed on Build 007 ---

X The web interface could now be accessed with http://localhost:2001/chat

X Replaced Inference Engine by the megahal engine (megahal.dll), as before
  evaluation occurs with <evaluate></evaluate> and learning with
  <train></train>.

  --- Installation Tips till a new cut of the setup program is made ---
  1) Add to the BOT.INI file the section [megahal] in BOT.XXX.
  2) unzip in directory .\dict the file megahal.zip.
  3) Ensure megahal.dll is on the same directory the EXEs are.


--- Fixed on Build 008 ---

X Integrated the MkDict 5th Order Markov engine for inference.
X <evaluate name="{MEGAHAL|MKDICT}"></evaluate>
X <train name="{MEGAHAL|MKDICT}"></evaluate>


--- Fixed on Build 009 ---

X Extend Verify method to recurse on <load>/<learn>, VERAIML.EXE now could
  parse an entire hierarchy of files if the top file is informed (startup.aiml).

X Review the consistent firing of the ONLLC event. Handler located in
  dev-init.aiml.

X Modified VERAIML adding -v switch to detect loops, -c switch is to verify
  integrity.
  Current format
           veraiml {-c} {-v} {-f filename}  (c & v are not mutually exclusive)

X Review the ill resolved tags during test sessions with the inference
  engine (AIML or bot?).

      --> Thanks for telling me, PEDROPEDROPEDRO.
          <getname/> tag defined three times on dev-setup.aiml, fixed

      --> A marsupial is ? is a marsupial is ?
          Current normalization doesn't remove ? but relies on a
          <pattern>* ?</pattern> to remove it, however this AIML based
          reduction fails on patterns like <pattern>abc *</pattern> because
          if the input matches "abc" the star absorbs the ? which remains
          in the input and get processed.
          Modified the forced normalization to avoid that.

X Added dev-looptest.aiml to introduce a loop reference case for VerAIML.

X Property JustifyOk on Bot Session (default FALSE) to generate or not
  the justification (avoid the performance problem when set to false).


--- Fixed on Build 010 ---

X Fixed bug on pattern matching when an empty tag defined as a new tag
  were used on places other than the 1st of the pattern.


X When a variable is filled with user input it's forced as Uppercase
  instead of preserving the initial case provided by the user.

X Input case is now preserved across the entire engine, either to set
  variables, store the bot activity, log, etc.

X Long chains of <li> have a performance problem (std-gender.aiml). Modified
  compiler directives to increase static heap allocated to the program
  (garbage collection performance issue??).
  Conditional solution --> the justification structure had been selectively
  overriden.

X AIML interpretation <gossip></gossip> previously nullify the evaluation,
  there are templates on the std set that requires the evaluation to be
  returned.


--- Fixed on Build 011 ---

X Fixed the ONRESET event generated by the bot when it's activated, now
  the outcome of the evaluation of the event is placed in the new property
  LastData, also this property is filled with the last answer of the bot
  at any moment.

X Method name ParseEntry changed into Reply (compatibility with mkdict/megahal)

X Changed the user interface of veraiml.exe ---> see veraiml.txt for details.

X The *LLC* string will not appear con "confused" (ONLLC) event evaluations;
  still is needed as a Think() on the ONLLC handler (dev-init.aiml) if
  veraiml.exe is executed with switch -v.

X (BIG) Change the criteria used to evaluate Word/*/_ now:
    - First _ is evaluated.
    - The Word is evaluated for a direct match.
    - The * is evaluated as the last resort.


--- Fixed on Build 012 ---

X Trap during load if a tag was used on <pattern></pattern> and doesn't exists.

X Compliance with AIML 1.0

   * Re-structure the bot setup AIML files, now

         dev-macros.aiml
         dev-aiml09.aiml
         dev-aiml10.aiml
         dev-botproperties.aiml
         dev-aimltransition.aiml   <--- garbage can of all limbo tags
         dev-aimlevents.aiml

     Also, changed startup.aiml, per-lola.aiml.

   * Reworked bot.ini file to define AIML 1.0 compliant properties.

   * Introduced empty tag <bot/> as native to the engine (previously were
     an alias of <get/>.

   * Changed base criteria on index variables from "0" to "1".

   * Tag <input/> now handled like an index variable (previously <clientsaid/>
   * Tag <that/>  now handled like an index variable (previously <botsaid/>.

   * Added to veraiml.exe the -n XX switch to correct (non looping)
     categories which recurse more than XX times.

   (see Pending for AIML 1.0 tags not yet supported).


--- Fixed on Build 013 ---

   Scientific minds beware of things that could not prooved conclusively
   as TRUE or FALSE.


--- Fixed on Build 014 ---

X Fix a bug when more than one <SRAI> were executed on the same template,
  the <STAR/> structure of the original pattern were contaminated by the
  <STAR/> values of the patterns involved on the evaluation of the <SRAI>;
  code made fully re-entrant now.

--- Fixed on Build 015 ---

X Reverted fix of <star/> evaluation to return " "+<star/>
X Ignoring <meta/> tag.

--- Fixed on Build 016 ---

X Modified the timer structure to detach it from the event framework.
  Now timers are set with <timer name="x" interval="nn">PATTERN</timer>
  being nn the interval in Secs. When the timer expires it resolves the
  pattern PATTERN as if it were processed by a <srai>PATTERN</srai>.

X Fixed several broken things on the original timer implementation.

X Added -q switch to veraiml, when activated (-s & -v MUST be activated too)
  will flag categories which are atomic at the top level (no star) but
  have one or more non-atomic categories at lower level. Will be signaled
  as "Fair --> Atomic+Star".

X Slight performance improvement, bogus object created and released
  during evaluation for no good.


--- Fixed on Build 017 ---

X When a new tag is defined with <define></define> if the new tag and it's
  mapped definition changes from tag to empty or empty to tag the attributes
  got lost.

     <set name="name"/>  --X--> <set><star/></set>             (incorrect)
     <set name="name"/>  -----> <set name="name"><star/></set> (correct)

--- Fixed on Build 018 ---

X Original input from user (non-normalized) made available by the engine
  on the variable "rawinput".

X OnVarChange system event fired when any variable is set to a different
  value, the event could be catch with the tag
       <onevent name="OnVarChange" value="varname"/>
  If no event handler is defined the event is just ignored.

X <thatstar index="X"/> Supported

X Eliminate the timer exception on Web Interface (on timeout).

X <gender></gender> thru macros on sys-macros.aiml

X Bogus additional space introduced on <that> evaluation.


--- Fixed on Build 019 ---

X Added switch "-t" to testbot.exe would make it to print a trace of the
  categories involved and the elapsed time of the evaluation (tracecat
  property of TAIMLBot object defined).

X Performance improvement:
   * Code profiled.
   * Most exercised structures on Pareto upper region 10% optimized.
   * Removed performance issue with 'justify' (JustifyOk property back to
     TRUE as default).

--- Fixed on Build 020 ---

X Added support for SOAP/XML P2P bot interchange (experimental) protocol.
   * <learn href="http:host:port">file</learn>
     ---> Remote load of aiml file
   * <srai href="http:host:port" alt="resp">pattern</srai>
     ---> Remote query for a given pattern {pattern}.

X Created BOTSERV.EXE (replacement for GUIBOT) with F.Piette TCP/IP objects.

X Fixed Justify structure (block on Web Interface).

X Modified Web Interface.


--- Fixed on Build 021 ---

X Introduced changes on AIML/SOAP/XML as per Jon Baer's note.

--- Fixed on Build 022 ---

X Introduced support for <aiml:request> inside a SOAP envelope.

X Cache of SOAP connections to preserve the sessionid.

X AIML 1.0 supported as <topicstar/> --> <get name="TOPIC"/>


X Correct dynamic AIML elaboration tag (<learn> variant) to stamp the
  source file correctly.

X Modify the <save/> command to save by file instead of topic.

  <topicstar index="X"/>

--- Fixed on Build 023 ---

X Introduced support for <system></system> (see tst-script.aiml).

X Introduced support for <script></script> tags.

X <script> == <javascript> == <vbscript> (see tst-script.aiml).

X Rework of Web interface HTML.

X Added <#COUNT> HTML meta-tag --> Number of current sessions active.

--- Fixed on Build 024 ---

X Changes on the XML/SOAP spect discussed with Jon Baer.

--- Fixed on Build 025 ---

X Changes after first XML/SOAP test

  * Made <aiml:topic/that/network> explicit.
  * Introduced a file serving POST function.

--- Fixed on Build 026 ---

X Added Category learning statistics during load on console version (TBot).
X Added configuration parameter [AIML].MERGE=(TRUE|FALSE) to control merging.

--- Fixed on Build 027 ---

X Load Performance Tunning with a code profiler

   * 64% of load time devoted to insert templates.
     ---> Hash table size optimization ---> now 3%
   * 12% of load time devoted to insert roots on Graphmaster.
     ---> Hash table size optimization ---> now 0.02%
   * 9.4% of load time devoted to insert chain of templates atoms.
     ---> Reserve a reference to last inserted to avoid search ---> now 0.001%
   *  8% of load time devoted to normalize spaces on templates
     ---> Assembly routine  ---> now 0.001%

   Load time theoretically in the range of 10% than previously
   Meassurement  Previous Avg(80 Cat/Sec) ---> Now Avg(500 Cat/Sec) (???)

X Parameter added [AIML].HASHSIZE (default 128000).
   * Hash table for templates.
   * 1/5 of the value for Hash table for roots on Graphmaster.


--- Fixed on Build 028 ---

X Change in the Std Library reqlib32 in the function IsNumeric screwed up
  the evaluation of <that index="2,n"/> generating a Conversion Exception
  because 2,n was equalled to an Integer.
  Restored the reqlib32 function and put some additional verification in
  place on the command processor.

--- Fixed on Build 029 ---

X JavaScript interpreter interface glitch (CSCRIPT.EXE) problems with
  version 5.1

--- Fixed on Build 030 ---

X Executing JavaScript scriptlets hang the machine on certain machines
  (Tom's test). Not related to the CSCRIPT.EXE version used (5.0 & 5.1)
  yield the same problem.
  Shelling out with a file interface instead of an internal pipe interface.

--- Fixed on Build 031 ---

X Exception handling while executing <system> tags.
X Fix System/Script lack of response when current path has spaces in the name.
X On SOAP requests method is added (/getAIML,/getQuery) based on the AIML
  tag that triggered the interaction (<learn>,<srai>).

--- Fixed on Build 032 ---

X Added support for MSXML3 XML Processor (XLS Processor).
X Added property [aiml].xsl to point the XSL file to use.
X Added Termination logic to TBot.

--- Fixed on Build 033 ---
X Small change in the pattern matching algorithm for patterns that contains
  tags to optimize the evaluation performance.

--- Fixed on Build 034 ---
X Added the display of the AIML file to trace messages (option -t)

--- Fixed on Build 035 ---
X Fixed bug on the <THAT> pattern matching.
X Added support for text to speech (TBot only)

     TBot -v n   where n=number of text processor to use [1..n]
                 (default 1)

--- Fixed on Build 036 ---
X Fixed speech DLL crash on termination (forced clean-up).
X Fixed initial bot result to be stored on <THAT/> (BotSaid stack).
X Added support for timers (TBot only), see tst-timer.aiml for test case.
     TBot -x     Will activate timers

--- Fixed on Build 037 ---
X Added Pattern text on Duplicate Pattern message during load.


**** VERSION 01.01.00 ****
X Re-Packaging of the Version.
    * Documentation.
    * Directory Structure of the distribution.


Pending
-------

? How to make the graphmaster double as the learning source for the inference
  engine? (that would make the answers from the inference engine consistent
  with those in the graphmaster).