[56] in Kerberos_V5_Development
minutes of meeting of 1990 January 16
jtkohl@ATHENA.MIT.EDU (jtkohl@ATHENA.MIT.EDU)
Tue Jan 16 16:56:53 1990
Minutes of 1990 January 16 meeting.
discussion of typedef's to structures vs. pointers to structures. For
opaque "objects" (e.g. caches), things will be typedef'ed to pointer to
struct. All others will be typedef'ed to structures and the programmer
will be responsible for using pointers where appropriate.
new types krb5_error_code and krb5_deltat will be added for error code
returns and time differences.
discussion of linking unnecessary libraries & functions, etc. when
using generic object-like interfaces (specifically libkt).
default kt lib will be a read-only interface; writers will need to
explicitly register to link in/use the read-write interface
flamage about atomicity, retries, error conditions on a changing key table,
etc. result: next_entry routine will give an error if the table
changed; the implementation is responsible for dealing with file open
modes and closing/opening if needed.
more flames about name parsing and quoting; end result: no change from
decision of earlier meeting.
Discussion about how to find kerberos servers for remote realms & how to
determine the realm of a specified service; no good answers exist for
now, and krb.conf/krb.realms are only barely tolerable solutions.
apply comments about linking in too much with libkt to the case of the
ccache
long-winded discussion of structure tags
discussion of which work we should farm out; concerns about time
constraints.
Action items:
-- Ken will check ANSI C spec about using _'s as leading chars for
identifiers.
-- Ken will propose specific places to put structure tags
-- Jay will edit libkt.spec to incorporate today's comments
-- Barry will edit librc.spec to incorporate today's comments
-- John will get back to external implementors with info about the
interfaces, and try to trade the replay cache for the file-based
credentials cache (so that the file-based cred cache is done at MIT)
-- Ilham will report back to the krb5dev discuss meeting with results of
his considerations of the aname/local name mapping issue (this
was from previous meeting, but accidentally omitted)
