J3/03-200

To: J3
From:  John Reid
Subject: Outcome of the meeting in Dresden
Date: 4 August 2003

We had a very successful meeting in Dresden, attended by 15 people,
representing 6 member bodies. This is an informal report. The actual
decisions are recorded in the resolutions (N1548) and the straw votes
are recorded in the draft minutes (N1549).

Most of the time was spent in considering the details of the draft new
standard. The only technical change made was very minor, in connection
with the syntax for the select type construct, where the examples were
not consistent with the bnf. WG5 decided that this was an error in the
bnf, not the examples, and decided to apply the changed bnf in two
other places, explicitly typed allocate statements and explicity typed
array constructors.

WG5 also considered a large number of editorial changes and made
recommendations to J3, which has responsibility for these.  Since
these will need to be considered by J3, we thought it best to make
them J3 papers as well as WG5 papers. I have put them all in

 ftp://ftp.numerical.rl.ac.uk/pub/j3

Mallory: please will you make each of these into a J3 paper? I have
also included here other relevant papers such as those concerning
interpretations, the TR, the resolutions, and the draft minutes.

WG5 decided that is will hold an informal letter ballot of its members
on the draft standard that is produced by the August meeting of J3,
prior to submitting it to SC22 for ballot as a final committee
draft. If all goes well, this ballot will commence in November and run
until March, which will allow WG5 and J3 to consider the result at
their joint meeting at the beginnning of May.

WG5 decided that the informal name of the language should be changed
to Fortran 2003 to reflect the fact that the technical content was
chosen in 2003. We forgot to consider if any edits to the draft
standard were needed, so I propose such edits below.

About a third of the meeting time was spent on the Technical Report on
enhanced module facilities and a new draft was constructed
(N1555). The following decisions were made

       a. Host association, as in the current draft, will be used.

       b. Complete agreement will be required between the interfaces
       as separately specified and as specified in the implementation
       (including recursiveness and purity, but not the name of the
       result variable).

       c. The implementation will be permitted to specify none of the
       interface.

       d. Each separate procedure will be individually labelled as
       such (rather than using any form of blocking or grouping as for
       CONTAINS). The label will be the token MODULE.

       e. The interface and its implementation will be permitted to be
       in the same program unit.

       f. There will be no limit on the number of generations (parents
       of parents).

       g. The IMPORT statement will not be permitted in a separate
       interface.

       h. The facilities will be compatible with Fortran 95, but edits
       will not be written against the Fortran 95 standard.

       i. The default implicit rules for a module procedure interface
       will be those of its host environment.

       j. WG5 was undecided whether the default implicit rules in a
       submodule should be those of its parent or implicit none and
       encourages J3 to consider this further.

WG5 asks J3 to consider the TR at its August and November meetings.
We will have a WG5 ballot after the November meeting and hope to send
the TR to SC22 before the end of December. This should allow the
result to be considered at the May meeting.

Early in the meeting, WG5 heard that four interpretations had passed
their J3 letter ballot and decided to accept three (99, 100, JP-17).
Some WG5 members had reservations about the fourth, JP-17, so this was
returned for further consideration. For details, see N1561.


EDITS

xiii:3-9. Change '2000' to '2003' five times.

4:26. Change '2000' to '2003'.