This archive is retained to ensure existing URLs remain functional. It will not contain any emails sent to this mailing list after July 1, 2024. For all messages, including those sent before and after this date, please visit the new location of the archive at https://mailman.ripe.net/archives/list/db-wg@ripe.net/
[db-wg] Proposal for returning IRT objects
- Previous message (by thread): [db-wg] Maintenance announcement
- Next message (by thread): [db-wg] Re: Deprecation of ip6.int scheduled for 1 June 2006
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Denis Walker
denis at ripe.net
Tue May 16 18:18:30 CEST 2006
[Apologies for duplicate mails]
Proposal for returning IRT objects
Motivation
----------
The original proposal for the IRT object (ripe-254) did not make adequate provision for presenting
the irt information to the user. A new query flag ("-c") was introduced. Because of the way this
flag was implemented and the objects it returned, it was not considered intuitive or helpful. It
can also cause confusion.
A discussion held at the RIPE 51 meeting decided that the original implementation was a
misunderstanding of the requirements. For more information, see AOB in the Database Working Group
minutes:
http://www.ripe.net/ripe/wg/db/minutes/ripe-51.html
There was further discussion at the RIPE 52 meeting, details of which can be found at:
http://www.ripe.net/ripe/maillists/archives/db-wg/2006/msg00052.html
Objectives
----------
- To increase the availability of the irt information to users.
- To promote the use of the IRT object.
- To make it easier for third party tool writers to find the correct contacts.
Background
----------
There has been confusion in the past about the implementation of the IRT object. We have been
studying the concept and looking deep into the whois query code. Our understanding of how this
works now, and how the community wants it to work in the future, has changed. This proposal
outlines the current behaviour and what we now understand to be the desired behaviour.
Current behaviour
-----------------
A standard query for address space returns the queried inetnum and related objects. This does
not include an IRT object, even if it is directly referenced by the queried INETNUM object.
By using the "-r" flag, the related personal objects are filtered out but ROUTE objects are still
returned.
By using the "-c" flag, an INETNUM object is returned with a referenced IRT object and objects
related to this INETNUM. This may not be the INETNUM in the query. No ROUTE objects are returned.
If no (less specific) INETNUM references an IRT object, nothing is returned.
By using the "-rc" flags, the related personal objects are filtered out, including the IRT object.
The only object returned is the referencing INETNUM object, which may not be the one in the query.
So it is not possible to find IRT objects without returning personal objects related to the
referencing INETNUM object.
New behaviour
-------------
A standard query for address space returns the queried INETNUM and related objects. This does not
include an IRT object, even if it is directly referenced by the queried INETNUM object. This is the
same as the current behaviour. We propose to change this default behaviour at the second stage of
implementation (see Implementation below).
By using the "-r" flag, the related personal objects are filtered out but ROUTE objects are still
returned.
A standard query for address space including the "-c" query flag returns the queried inetnum and
related objects. This will also include an IRT object referenced by the queried INETNUM object or
the closest less specific INETNUM object, if one exists. It will also still include any ROUTE objects.
By using the "-rc" flags, the related personal objects are filtered out. The IRT object will still
be returned along with any ROUTE objects.
When using any other address space query flags (lLmMxb) or querying for other object types, query
behaviour remains the same as the current behaviour. If the "-c" query flag is used, it will be ignored.
Implementation
--------------
We propose a two stage implementation. The majority of the address space queries are made without
any specific flags. We therefore propose to change the behaviour to return an IRT object by default
(if one exists) only at the second stage.
For the first stage, we propose changing the behaviour of the "-c" flag and the "-rc" flag
combination as described above. This should not make any fundamental changes to the way the queries
work. Existing tools and scripts should not fail. The web query form on www.ripe.net will add the
"-c" flag to address space queries, making it a default. This only occurs if no other address space
query flags (lLmMxbc) are present.
At the second stage, the query server will add the "-c" flag to any address space query that does not
include any of the address space query flags (lLmMxbc). This will make standard queries return IRT
objects by default.
This second stage changes the current behaviour of queries. This may break some tools and scripts.
We can mitigate this by providing new query flags or using existing ones in different ways so that
the behaviour resulting from the first stage of implementation can be reproduced.
Tool writers
------------
Currently any third party tools that use the "-c" flag to find IRT objects run the risk of being
blocked for returning too many personal objects. With the new behaviour, the "-rc" flags can be used
to return the IRT objects without the personal objects. If only the IRT object is needed,
"-rc -T irt" can be used.
Notes
-----
Any references to INETNUM and ROUTE objects also imply INET6NUM and ROUTE6.
A consequence of this change is the return of an object with no direct reference from the queried
object.
With grouping turned on (default without "-G"), we assume that the IRT object should be grouped with
the queried INETNUM object.
--
Denis Walker
Software Engineering Department
RIPE NCC
- Previous message (by thread): [db-wg] Maintenance announcement
- Next message (by thread): [db-wg] Re: Deprecation of ip6.int scheduled for 1 June 2006
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
[ db-wg Archives ]