At 01:01 PM 11/30/98 +0100, Joao Luis Silva Damas wrote: Yes, that may be all correct - from a techno standpoint. What I am stressing is the human engineering aspect. Up till now, thousands of hits are sent to the whois.ripe.net every month to determine whether a domain exists or not. Users have been doing this for years. Suddenly, the output is different and they do not realize it. I highly doubt that the human engineering aspect was discussed or analyzed to any extent in the DB WG. -Hank
Dear all,
The implementation of the referral mechanism is a new feature of the latest release of the RIPE DB software.
The design of this feature was specified some time ago by Carol Orange and Wilfried Woeber (please see http://www.ripe.net/meetings/ripe/ripe-27/pres/rout e/referral/) and discussed in the RIPE db-wg mailing list (please see http://www.ripe.net/mail-archives/db-wg/19970501-19970601/threads.html)
According to this proposal the algorithm to be implemented is
(extracted from the proposal): *********** Algorithm --------- If the following query is submitted to the RIPE database:
whois [flags] xxx.yyy.zz
the algorithm will work as follows:
------- NAME="xxx.yyy.zz"
until (NAME = "") { If object with domain = NAME found, If object contains referral (see "Referral" below) forward query (see "Forward" below) Else return object Else Strip(NAME) (xxx.yyy.zz -> yyy.zz, etc) }
NOTES ----- 1. We move up the tree here, and return the next level answer if present. If the query requests "tuintje.cwi.nl" (Piet, do you mind?), they currently get "No entries found ...". In the new mechanism, they would get:
domain: cwi.nl . . . mnt-by: NL-DOMREG changed: hostmaster@domain-registry.nl 19950227 source: RIPE
Moreover, if the object for cwi.nl contains a referral, the query would be passed on to the specified server as explained below.
2. The algorithm will be set off by any query which causes a search in the domain object index. This means any query with "-T domain", or any general query (no -T flag) with something that "looks like a domain". *************
Currently and due to s small desynchronization within the RIPE NCC DB group and the fact that the proposal was not clear about a particular point the referred query is for the truncated object in the RIPE DB containing the refer attribute. We want to modify this so that the query to the authoritative server is for the user's initial query and not the result of the truncating algorithm. This change will take place either today or tomorrow and a note will be sent to the db-wg.
As a side note I would like to stress that the RIPE Database is NOT the source of authoritative data for domain objects (although we are happy to provide a repository for TLD administrators). This is UNLIKE the internic whose business is to sell names.
The RIPE NCC is therefore acting according to users directives that were discussed and agreed on an open way.
We are happy to hear any comments about this or any other issues regarding the DB software and will keep working to suggest and provide users with functionality that will help them the most.
Please send any comments to the RIPE mailing lists.
Best regards, Joao Damas RIPE DB Group manager RIPE NCC
NCC local-ir list Moderator <lir-mod@ripe.net> writes: * * I feel this is more appropriate to be discussed on the db-wg. * * Mirjam Kuehne * RIPE NCC * * ------- Forwarded Message * * Date: Thu, 26 Nov 1998 19:21:13 +0200 * To: <local-ir@ripe.net> * From: Hank Nussbacher <hank@ibm.net.il> * Subject: Ripe change to whois * * I may have missed it, but a serious *bad* change has occured to the whois * interface. Previously, if I did a check on abcdef.co.il and abcdef did not * exist, whois would return an error message. Now, whois returns the domain * and role records for the next upper level domain - i.e. co.il or .il. Other * examples: zzzz.ac.uk, zzzz.ac.at. * * How did I find out about this? I received 3 phone calls already in the past * 24 hours from users in Israel stating that I personally am registered as * owner of hundreds of domains. They state that being on the board of the * ISOC-IL which oversees the registration, I have violated the rules (limit of * 10 domains per person - other countries have similar restrictions). I have * tried explaining to them that I am not the owner of these domains. If it * was 1 user, I would assume they didn't read the entry correctly. But the * fact that 3 totally different people have called me to ask whether I would * sell certain domain names to them, I can only conclude that the RIPE change * is not that successful. * People have been used to getting an error message from RIPE, similar to the * "No match" one gets from Internic. When they now get a valid response, they * don't bother to check that the response is not for the domain they * requested. There is no error message or error code. * * This needs to be rectified (and changes discussed in the future - ignore if * it has been mentioned). * * - -Hank Nussbacher * ISOC-IL board member * * * * * ------- End of Forwarded Message * * *
Dear Hank, May be (or may be not) the "human engineering aspects" were discussed enough but, please, if things like this affect your work so much, get into the db-wg mailing list and give your input. The TLD wg was also involved in this and they were actually very much looking forward to have this mechanism in place. If you still feel that "This needs to be rectified" you can discuss this with the community, although it would have been better to discuss all this when the design was proposed instead of going around the loop again (which, of course, we'll do if users want to). I'll wait and listen. Regards, Joao Hank Nussbacher <hank@ibm.net.il> writes: * At 01:01 PM 11/30/98 +0100, Joao Luis Silva Damas wrote: * * Yes, that may be all correct - from a techno standpoint. What I am * stressing is the human engineering aspect. Up till now, thousands of hits * are sent to the whois.ripe.net every month to determine whether a domain * exists or not. Users have been doing this for years. Suddenly, the output * is different and they do not realize it. I highly doubt that the human * engineering aspect was discussed or analyzed to any extent in the DB WG. -H * ank *
On Mon, Nov 30, 1998 at 02:17:54PM +0100, Joao Luis Silva Damas wrote: Hi Joao,
If you still feel that "This needs to be rectified" you can discuss this with the community, although it would have been better to discuss all this when the design was proposed instead of going around the loop again (which, of course, we'll do if users want to).
I'm not sure what needs to be 'rectified'. But if you silently change the semantics of a public database more than a year after a discussion that wasn't attended by its current users, then you shouldn't wonder about complaints. But complaining alone doesn't achieve anything. As it appears the database lacks the feature of an exact-match query and I kindly propose to add such a feature. Regards, -- i.A. Michael van Elst / phone: +49 721 6635 330 Xlink - Network Information Centre \/ fax: +49 721 6635 349 Vincenz-Priessnitz-Str. 3 /\ link http://nic.xlink.net/ D-76131 Karlsruhe, Germany /_______ email: hostmaster@xlink.net [ Xlink Internet Consulting GmbH, Sitz Koeln ] [ Amtsgericht Koeln HRB 3526, Geschaeftsfuehrer: Michael Rotert ]
Hi, Michael van Elst <mlelstv@xlink.net> writes: * * I'm not sure what needs to be 'rectified'. But if you silently change * the semantics of a public database more than a year after a discussion * that wasn't attended by its current users, then you shouldn't wonder * about complaints. Well, Hank is hardly a new user. And yes, we took too long to deploy this feature, however, I can't see that the issues would have been different a year ago (maybe I am wrong). I am not having an exercise on defending against changing this. I am more than glad to make changes if the community asks for them (and to propose new ones that we think make sense). I want to see what the rest of the people think since the previous consensus was different to what I am seeing now. * * But complaining alone doesn't achieve anything. As it appears * the database lacks the feature of an exact-match query and I * kindly propose to add such a feature. * Well the purpose of the design is to give the user information about the parent domain if a certain domain is missing. That's also how the rest of the lookups in the RIPE DB work and is also what you do when looking in the DNS for contact info for a domain and you can't find it. May be people now see the need for an exact matching option (there is already one to disable referral: -R). However if you make this the default behaviour then the purpose of the referral mechanism is defeated. Joao * * Regards, * -- * i.A. Michael van Elst / phone: +49 721 6635 330 * Xlink - Network Information Centre \/ fax: +49 721 6635 349 * Vincenz-Priessnitz-Str. 3 /\ link http://nic.xlink.net/ * D-76131 Karlsruhe, Germany /_______ email: hostmaster@xlink.net * [ Xlink Internet Consulting GmbH, Sitz Koeln ] * [ Amtsgericht Koeln HRB 3526, Geschaeftsfuehrer: Michael Rotert ] * *
On Mon, Nov 30, 1998 at 04:58:09PM +0100, Joao Luis Silva Damas wrote:
Hi,
Hi Joao,
Michael van Elst <mlelstv@xlink.net> writes: * I'm not sure what needs to be 'rectified'. But if you silently change * the semantics of a public database more than a year after a discussion * that wasn't attended by its current users, then you shouldn't wonder * about complaints.
Well, Hank is hardly a new user. And yes, we took too long to deploy this feature, however, I can't see that the issues would have been different a year ago (maybe I am wrong).
I'm just saying that there was some lack of communication. I think the RIPE database is too important to many people to modify it as silently as in this case. The last changes (suppressing of 'changed' lines and expansion of 'route' (?) objects) had some discussion immediately before they were implemented in the production system. I'd have preferred an more explicit announcement of changes, especially when semantics are changed.
* * But complaining alone doesn't achieve anything. As it appears * the database lacks the feature of an exact-match query and I * kindly propose to add such a feature. *
Well the purpose of the design is to give the user information about the parent domain if a certain domain is missing. That's also how the rest of the lookups in the RIPE DB work and is also what you do when looking in the DNS for contact info for a domain and you can't find it.
The point is that it is incompatible. I don't mind having a 'less-specific' query (although I would have made this an explicit option and a bit more orthogonal). It is also not exactly like the rest of the lookups. There are hierarchical objects (inetnum, route) and there are more individual objects (like persons) where query types being 'less-specific' or 'more-specific' do not make sense. You also cannot do a more-specific search for domains (nor would it work well seeing how flat the domain tree is).
May be people now see the need for an exact matching option (there is already one to disable referral: -R). However if you make this the default behaviour then the purpose of the referral mechanism is defeated.
Since I like orthogonal designs I suggest to add an -X option to return only exact matches for any kind of object. Regards, -- i.A. Michael van Elst / phone: +49 721 6635 330 Xlink - Network Information Centre \/ fax: +49 721 6635 349 Vincenz-Priessnitz-Str. 3 /\ link http://nic.xlink.net/ D-76131 Karlsruhe, Germany /_______ email: hostmaster@xlink.net [ Xlink Internet Consulting GmbH, Sitz Koeln ] [ Amtsgericht Koeln HRB 3526, Geschaeftsfuehrer: Michael Rotert ]
Hello Joao. [snip]
Well the purpose of the design is to give the user information about the parent domain if a certain domain is missing. That's also how the rest of the lookups in the RIPE DB work and is also what you do when looking in the DNS for contact info for a domain and you can't find it.
Yes, sure, if I do a lookup for a single IP address I am more than happy to get shown the next inetnum segment. But for domains it fails to make much sense for me. How rare are the cases in which you have to look up the domain object of de? And if you need to, I cannot help but ask why it is not done by hand. inetnum and domain objects are rarely mixable when speaking of hierarchical systems... it does not scale equally, I am afraid. I would like to know some reason for this change by the tld workgroup.
May be people now see the need for an exact matching option (there is already one to disable referral: -R). However if you make this the default behaviour then the purpose of the referral mechanism is defeated.
Let us think the other way round. Make it an whois option. Whatever, whois -L nothing.de works fine enough for me, but it is been all but nice hearing from users something weird happened about domains (and they were not even using the RIPE DB itself). Nicer If I had found out reading the announcement, nothing more. Regards, Alexander Koch -- SGH Internet Division, Alexander Koch, Systems Administration Hannover, Germany, Phone +49 511 909198 0, Fax +49 511 391307
Hi again, I am hearing complaints about lack of communication regarding this change in the DB. I would like to point out the following: We try to announce things as much as possible and are always looking for suggestions and ways to improve this. However, it is and will be impossible to reach all end users (we don't want to start spamming, do we?) We expect to reach people whose interaction with the DB is a bit stronger than the occasional end user. For this we have the mailing lists, the RIPE meetings and the Web site. Discussions about developments take place in RIPE meetings and mailing lists. For the people that can't attend RIPE meetings all presentations are available at the Web site. This has been so for a some time. If you are not subscribed to mailing to mailing lists there is not much we can do, other than suggest you to subscribe if you rely on the RIPE DB for your work. I don't think we should be sending mails with announcements to everyone registered in the RIPE DB. Now, in a previous mail I pointed out a few URLs regarding discussions of the referral mechanism. These, in my view, are hardly 4 light years away (more like a click away) and they are at a place where people interested are supposed to look anyway. Not only this, but discussions have been going on since then and as you can see in the archives (publicly available at our web site) details regarding the referral mechanism were being discussed in the tld-wg and db-wg lists as recently as October 6th 1998 when we were finishing the coding and changes could still be easily introduced (as they were). If people don't want to be subscribed to these lists but the work going on there is relevant to them may be it would be a good idea to, at least, make it a monthly task to overview the archives and look at the presentations given during RIPE meetings. Otherwise, how can we make sure that we reach all interested parties without spamming the rest? So, although we may have room for improvement in getting announcements to a wider audience (and we will do our best), in this particular case and the people involved, the excuse of not knowing anything about what was going on is hardly one at all. As to why the referral as is implemented makes sense or not, I will leave this to the tld community, as the interested parties, to explain. And, I want to stress this again, everything is open to change if we a consensus is reached on what is needed/wanted. Regards, Joao Alexander Koch <akoch@sgh-net.de> writes: * Hello Joao. * * [snip] * > Well the purpose of the design is to give the user information about the * > parent domain if a certain domain is missing. That's also how the rest of * the * > lookups in the RIPE DB work and is also what you do when looking in the DN * S * > for contact info for a domain and you can't find it. * * Yes, sure, if I do a lookup for a single IP address I am more than happy * to get shown the next inetnum segment. * But for domains it fails to make much sense for me. How rare are the * cases in which you have to look up the domain object of de? And if you * need to, I cannot help but ask why it is not done by hand. * * inetnum and domain objects are rarely mixable when speaking of * hierarchical systems... it does not scale equally, I am afraid. * * I would like to know some reason for this change by the tld workgroup. * * > May be people now see the need for an exact matching option (there is alre * ady * > one to disable referral: -R). However if you make this the default behavio * ur * > then the purpose of the referral mechanism is defeated. * * Let us think the other way round. Make it an whois option. * Whatever, whois -L nothing.de works fine enough for me, but * it is been all but nice hearing from users something weird * happened about domains (and they were not even using the RIPE * DB itself). Nicer If I had found out reading the announcement, * nothing more. * * Regards, * Alexander Koch * * -- * SGH Internet Division, Alexander Koch, Systems Administration * Hannover, Germany, Phone +49 511 909198 0, Fax +49 511 391307 *
On Tue, Dec 01, 1998 at 12:42:52PM +0100, Joao Luis Silva Damas wrote:
Hi again,
Hi Joao,
I am hearing complaints about lack of communication regarding this change in the DB. I would like to point out the following:
[...]
So, although we may have room for improvement in getting announcements to a wider audience (and we will do our best), in this particular case and the people involved, the excuse of not knowing anything about what was going on is hardly one at all.
yeah, it's the victims fault all over again. Sorry, this 'discussion' is only wasting time. If you don't see that this sudden modification causes problems then I cannot help you. Regards, -- i.A. Michael van Elst / phone: +49 721 6635 330 Xlink - Network Information Centre \/ fax: +49 721 6635 349 Vincenz-Priessnitz-Str. 3 /\ link http://nic.xlink.net/ D-76131 Karlsruhe, Germany /_______ email: hostmaster@xlink.net [ Xlink Internet Consulting GmbH, Sitz Koeln ] [ Amtsgericht Koeln HRB 3526, Geschaeftsfuehrer: Michael Rotert ]
participants (4)
-
Alexander Koch
-
Hank Nussbacher
-
Joao Luis Silva Damas
-
Michael van Elst