[R6RS] my notes on today's conference call (29 March 2006)

[William D Clinger]

Conference call March 29 2006 starting at 12:45pm
Kent, Mike, Anton, Will (Matthew arrived about 12:59pm)

0. finalize agenda (1 minute)
    consider i/o before macros so Matthew will be here for macros

1. action items from 3/22/2006 (5 minutes)
   - look over proposed condition hierarchy and details (All)
        carried over
   - draft syntax-case SRFI (Kent) [by 3/29]
        drafted and emailed, but not yet checked into Subversion
   - propose a set of symbol characters that must be hex-encoded (Flatt)
        see revised Unicode SRFI

2. Daylight-Savings/Summer-Time meeting time (2 minutes)
    go with 12:45pm EDT on Wednesdays

3. core/base/non-base-library split (10 minutes)
    Anton posted a proposal
        SRFI 83 "scheme://r6rs" doesn't allow smaller or larger languages
            Anton proposed S-expression-based syntax
            what matters is not syntax, but what the language can be
    Anton, Kent, and Matthew will propose a syntax
        that implies the lexical syntax of the library form,
        and uses imported libraries for everything else
    Anton will create strawman list of proposed libraries

4. continue I/O discussion (15 minutes)
    Mike: major issue is with text encoding
        text encoding is binary data to binary data
        general agreement on that, with some uneasiness
    what are descriptors?
        completely implementation-dependent
    file-options, buffer modes, eol styles, etc
        some allowed to be inexact reals
        some required to be disjoint types
        why not use symbols?
        let's choose some mechanism and use it consistently:
            symbols, lists of symbols, or some new enumeration mechanism
    u8vector and blobs
        allowing them to be the same, but not requiring them to be,
            invites portability problems
        make them the same, but eliminate the u8vector name
            u8vector is misleading because it promotes one view
                (as unsigned 8-bit bytes) at expense of other views
        blobs has database precedent (binary large object)
            but ugly, undignified, and unsuggestive
        possible alternative names for u8vector-ref:
            bytes-ref
            binary-vector-ref
            bvector-ref
            bv-ref
    if cond's => hack is the main argument for returning #f at eof,
        then why not add an eof=> hack?

5. syntax-case macros (20 minutes)
    deferred to next week

6. adjourned about 1:43pm

Will