[Date Prev][Date Next] [Chronological] [Thread] [Top]

Re: New schema

At 09:43 AM 4/26/99 +0200, Julio Sánchez Fernández wrote:
>
>Tonight (CET) I will start committing parts of the new schema to the repository.

Great.

I'd very much like to see
>I must say this has become a *very* frustrating process.  Despite what one
>might think, RFC1274 does *not* describe the object classes and attribute
>types that we had with names including 'pilot' and such.  They are very
>similar, but not the same.  The definitions Netscape DS use are very similar to
>ours, with very small deviations.  I can only deduce that the pilot schema
>evolved after RFC1274 and Umich, and later Netscape, tracked that evolution.
>That, or Umich literally made up the changes and were inherited by Netscape.
>Anyway, I have been unable to find updated COSINE pilot schemas that reflect
>those changes.  As a consequence, some objects from the pilot do not have
>OIDs known to me.  It is little consolation that Netscape does not have them
>either.

I'm sure someone on LDAPext would know.

>On the other hand, I have found two large blocks of definitions that do not
>come from RFCs as far as I know.  One was expected: the Umich definitions.

We should move away from Umich in favor of newer, better defined, schema
items.  We should be able to eliminate our dependencies upon these items.

>The second is something I did not even knew was there:  US-centric things,
>many of which have names that include either 'nadf' or 'fips'.  I do not
>know where they come from.  I am desperate.

If we don't need these.  I recommend we isolate them so that they can
be removed from the distribution before release.

>Well, since renaming files in CVS is messy, I will announce what files I will
>create tonight so that objections can be raised before I do.  These files will
>be added to servers/slapd:

How about we add them to servers/slapd/schema?

>	- slapd.std.schema:  Definitions from RFC2252 and RFC2256 in the new
>		syntax
>	- slapd.pilot.schema:  Same for RFC1274 ammended as deemed necessary,
>		this is incomplete and needs revision.
>	- slapd.umich.schema:  Same for Umich definitions, when available
>	- slapd.misc.schema:  Same for assorted definitions, like RFC2247,
>		RFC2377 and, possibly, RFC2307.  In general, final destination
>		for all definitions not included in other files.
>	- slapd.other.schema:  Temporary file in the old format for all
>		definitions for which adequate ASN.1 or RFC2252-style definitions
>		are not known.  Hopefully, this file will go away.

I'm am not sure how best to organize the information in files.
I do think we should separate items by source.  Hence, I think
we should have a separate file per source:

Standard track RFCs:
	rfc2252.schema	(required)
	rfc2256.schema	(x500 user)

Informational/Experimental RFCs:
	rfc1274.schema	(pilot)
	rfc2247.schema	(dir-enabled apps, info)
	rfc2307.schema	(nis, experimental)
	rfc2377.schema	(internet naming)

IETF drafts:
	inetorgperson.schema	(latest draft, optional)

Other:
	rfc1274.schema		(rfc1274 items)
	rfc1274++.schema		(enhanced rfc1274 items)
	ol-slapd.schema		(OpenLDAP slapd items)
	ol-APPLICATION.schema	(OpenLDAP schema for 'APPLICATION')

Deprecated:
	umich.schema		(umich items)
	deprecated.schema	(deprecated schema items)


OpenLDAP could, by default, install all schema files in the
$sharedir/openldap (ie: /usr/local/share/openldap/schema).
We can create a master.schema file that looks like:
	#
	# master schema file for OpenLDAP slapd
	#

	# Required
	include %SCHEMA%/rfc2252.schema
	include %SCHEMA%/rfc2256.schema
	include %SCHEMA%/ol-slapd.schema

	# Recommended
	include %SCHEMA%/rfc1274++.schema
	include %SCHEMA%/rfc2377.schema
	include %SCHEMA%/rfc2247.schema

	# Optional
	include %SCHEMA%/rfc2307.schema
	include %SCHEMA%/ol-APPLICATION.schema

	# Deprecated
	# include	%SCHEMA%/umich.schema
	# include	%SCHEMA%/deprecated.schema

Of course, an default installed schema should be a subset
of the above.  I think we should include schema only as required
by standard or by software within the distribution.  Software
in the distribution should be updated to use latest IESG/IETF
defined schema items.

RFC1274 items concern me greatly.  It sounds like the RFC
needs to be updated.  This would be a good topic for LDAPext.
We probably should steer widely around RFC1274 as possible.
Hence, maybe rfc1274++.schema should be limited to only those
items to which we really need.   Hence:
	rfc1274.schema (as defined in the RFC)
	rfc1274++.schema (common use 'pilot' items)
	rfc1274opt.schema (optional, common use 'pilot' items)

We should consider moving towards better, well-defined (RFC or near
lastcall IETF draft) replacement schema items (where they exist).

Kurt