As no one pointed out any problems, these are unchanged from the draft
minutes sent out in August.
Chris
Minutes of the 8/13/97 meeting of the Schema Listing BOF
Special thanks to John Strassner for taking the minutes!
The agenda was as follows
1: Bash the charter
2: Current holes in the Schema Listing Requirements document (Chris
Apple)
3: Discussion of open issues
Discussion:
1: There were no problems with the charter
2: Holes in the document that need to be patched
- doesn't match Scott Bradner's requirement terminology
- Security considerations
- The philosophy of unlimited read and tightly controlled write needs to
be spelled
out in the document
3: Other issues resolved
- Syntax specs should be BNF
- Localized versions of the schema should be made available and should
be pointed
to by the main schema. This includes language tags.
- OIDs will be used to uniquely identify schema
- A question about two competing requests (i.e. same names, but
different content)
came up. This should be handled in a separate process document.
However, since
the service will be assigning OIDs to the schema, each schema will
have a unique
name.
- Does a change in the machine-readable portion of a schema require a
rev of the
entire schema? YES.
- The schema listing file name for a given schema will be the last
component of the
OID assigned to that schema. We should not include any other human
semantics
in the file name
- Again, this is a listing service, not a vetting service. Schema
standardization will
happen after schema are listed
- We will be adding a usage scenario section to the document.
- We will be selecting one syntax as a required syntax for the service
and will add
two more if we can get the parsers. The three agreed on by the group
are LDIF,
ASN.1, and XML. We'll do a small list poll for the required one.
- We will be requiring FTP and HTTP access to the repository.
4: Open issues for the list
- Should any document that specifies schema be allowed to be registered?
This is
an issue for the list
- Who owns the database? Who runs it? This will be discussed as we get
closer to
the process document.
- Do we need to track and forbid display name collision of schema and
attributes?
I.e. two people trying to register the 'Person' schema even though
they have
different OIDs?
- Should the listing service be searchable? We need to decide whether it
is, and if
so, what engine would be used. There are also the matters of UI and
complexity
of the interface.
- Can this service be tied into the URN infrastructure?
- For Jan 1, 1998, should we allow machine submission of schema. We are
leaning
towards NO but will take to the list.
- Which (if any) of the submitted schema will be made into RFCs?
- Should schema be signed so that people can verify that their schema
are indeed
the ones that they sent? Which signature algorithm should be used?
- Do we store metaschema (i.e. schema not tied to any protocol, such as
the Dublin
Core). If so, how? How do we tie them to their expressions in various
protocols?
- What character sets should we support? How do we support them?
- What are the required metadata elements for a schema listing?
- Should we push for LDAP and SMTP access to the repository for Jan 1?
- How do we control versioning?
- Can we provide an automated submission process for Jan 1?
- Is the store moderated to avoid BadGuy(tm) or StupidGuy(tm) attacks?
One general
rule is that we could agree that unsigned schema are automatically
rejected
5: Action items
a: change the acronym of the working group so that it doesn't contain
the name
schema...
b: Put together the engineering team. Volunteers were:
Michael Mealling - NSI
Sam Sun - CNRI
Mark Wahl - Critical Angle
Sanjay Jain - Oracle
John Strassner - Cisco
Chris Weider - Microsoft
c: rev the document and start the list discussions. We will cut off list
discussion
no later than 15 September to insure that the engineering team has
everything it
needs.
