[R6RS] my notes on today's conference call (24 August 2006)

[William D Clinger]

August 24 2006 8:00am-10:00am EDT
All present by 8:03am EDT:
Kent, Will, Mike, Anton, Matthew

0. finalize agenda (1 minute)

1. action items from 8/22/2006 (5 minutes)
   - revise I/O port SRFI with read-time transcoder arguments (Will)
        done
   - carried over:
     - revise safety paragraphs (Will)
        done
   - note: reserve last week of August (THIS WEEK) for proofreading

2. Report status (5 minutes)
   - items ready for review
     <item>                    <editor>   <to be completed by>   <done?>
     unicode                   Matthew    Aug. 03                done.
     quasiquote                Kent       Aug. 08                done.
     letrec*                   Kent       Aug. 08                done.
     let-values & let*-values  Kent       Aug. 08                done.
     records                   Kent       Aug. 24
     enumerations              Will       Jul. 12                done.
     bytes                     Will       Aug. 03                done.
     equal?                    Will       Aug. 01                done.
     list-library              Anton      Aug. 21                done?
     case-lambda               Anton      Aug. 21                done?
     when & unless             Anton      Aug. 08                done.
     arithmetic                Will       Aug. 22
     set-car!, set-cdr!        Anton      Aug. 21                done?
     condition-type details    Anton      Aug. 21
     dynamic-wind              Kent       Aug. 15                done.
     read-syntax, syn. forms   Matthew    Aug. 20                done.
     hash tables               Anton      Aug. 24
     safe/unsafe mode          Kent       Aug. ?? (waiting for Will)
   - items not ready for review
     core/library split
     exceptions & conditions (done; reviewed by Matthew weeks ago)
     eval
     I/O
     hash tables
     syntax-case
     library system
     scripts
     r5rs compatibility library
     operational semantics

3. I/O (20 minutes)
   - accept transcoder change?
     (see srfi/io/port-io-alt.html)
        need to query transcoder from stdio, therefore all ports
        (define (read-char port)
          (get-char port (port-locale-transducer port)))
            Yes: Will, Kent, 
            No: Matthew, Mike
            Abstain: Anton
        Matthew moved we accept the transcoder change as amended above.
        Will seconded.
            Yes: Will, Kent, Matthew
            No: 
            Abstain: Mike, Anton
   - what happens with illegal byte sequences during textual i/o?
        Mike's proposal:
            raise a specific continuable exception (default)
            except specific transcoders can specify
                ignoring or replacement semantics
            See email of 21 July 2006
        Will moves we accept Mike's proposal of 21 July
            with semantics of the continuation to be worked out.
        Mike seconded.
            Yes: Will, Kent, Matthew, Anton, Mike
            No: 
   - flush or keep update-transcoder?
        Mike moves we add accessors and pin down eqv? semantics.
        Will seconded.
            Yes: unanimous
        Kent moved we flush update-transcoder and make
            transcoder into a three-argument procedure
            make-transcoder with three arguments (two
            optional): codec, eol, errors.
        Mike seconded.
            yes: Mike, Kent, Anton
            no: 
            abstain: Will, Matthew
   - can lookahead-u8 or get-u8 return something other than eof if
     the last operation on the port was a lookahead-u8 that returned
     eof but the underlying file/object was extended in the meantime?
        Kent moved that read-char may return end-of-file
            and then return something else.
        Will seconded
            yes: Kent, Matthew, Anton, Mike
            no: 
            abstain: Will
        Kent moved that if lookahead-whatever returns
            an end-of-file object, then the next call
            must also return end-of-file.
        Will seconded
            yes: Matthew, Mike, Kent
            no: 
            abstain: Will, Anton
        Kent moved that if an exception is raised while reading
            a multibyte character because the character is
            incomplete, a subsequent call is not *required* to
            raise an exception
        Will seconded
            yes: Will, Kent, Matthew, Anton
            no: 
            abstain: Mike
   - must lookahead-u8 or get-u8 return eof if the last operation on the
     port was a lookahead-u8 that returned eof but the underlying
     file/object was extended in the meantime?
        yes; see above
   - must lookahead-u8 or get-u8 raise an excption if the last operation
     on the port was a lookahead-u8 that raised an exception due to
     insufficient bytes but the underlying file/object was extended in the
     meantime?
        no; see above
   - add clear-output-port?
   - add clear-input-port (possibly as flush-input-port)?
        no one much cares, and more important issues await
   - should get-line recognize the Unicode line separator in adition to
     the specified or default end-of-line convention?  Should it recognize
     all common conventions even if they aren't the current one?
        Will moved that get-line be required to recognize
            line separator in addition to what eol-style is
            specified by the transcoder
        Mike seconded
        yes: Will, Kent, Matthew, Anton, Mike
        no: 
        abstain: 
        Will moved that get-line be allowed to recognize
            other conventions in addition to the eol-style
            specified by the transcoder
        Motion fails for lack of a second
   - do we want to translate different end-of-line conventions to linefeed?
        yes
   - add open-bytes-output-port, get-output-bytes, open-string-output-port,
     get-output-string?
        Kent moved we add them
        Will seconded
            motion passes without objection
        Does get-output-string clear the output?
            No, it does not implicitly clear the output.
            Kent will propose a new procedure that clears the output.

4. Core/library split (20 minutes)
   - number and size of libraries to be described by R6RS
   - size of "mimimal language" presented by document
   - congruence between R6RS document and R6RS libraries
        general discussion of above issues;
        we have only 1 week left
   - hierarchies and something like Java's wild card syntax?
   - do wildcards apply outside of r6rs hierarchy?

5. adjourned at 10:05am EDT