From oli at isnic.is Mon Nov 26 16:34:38 2001 From: oli at isnic.is (Olafur Osvaldsson) Date: Mon, 26 Nov 2001 15:34:38 +0000 Subject: host tool used for RIPE hostcounts Message-ID: <20011126153438.G43008@isnic.is> Hi, As I understand, you made the host tool I'm referring to. We've been using the host tool from ftp.ripe.net for the hostcounts since I don't remember when but noticed some strange behaviour this month. We got a bunch of these errors: *** incomplete resource record for ppp129-110.islandssimi.is, offset 8186 ;; response truncated to 8192 bytes ;; got answer, 23618 bytes: ;; ns_initparse: Message too long *** expand error in resource record for islandssimi.is, offset 8186 After some diging it looks like this only happens when transfering from nameservers with version 9 of bind, wich seems to be sending the whole transfer at once instead of sending one RR at a time like bind 8 does. I've fixed this by patching host.h and set the MAXPACKET size to what I think is the maximum in bind 9. There is probably a better solution to this but this atleast makes things work correctly for now. I've Cc'd this to the dns-wg at ripe.net as there are probably alot of people there geting into the same trouble. Oli -- Olafur Osvaldsson Systems Administrator Internet Iceland inc. Tel: +354 525-5291 Email: oli at isnic.is -------------- next part -------------- *** host.h.orig Mon Nov 26 15:14:45 2001 --- host.h Mon Nov 26 14:39:12 2001 *************** *** 68,77 **** #define BROADCAST_ADDR ((ipaddr_t)0xffffffff) #define LOCALHOST_ADDR ((ipaddr_t)0x7f000001) ! #if PACKETSZ > 8192 ! #define MAXPACKET PACKETSZ /* PACKETSZ should be the max udp size (512) */ #else ! #define MAXPACKET 8192 /* but tcp packets can be considerably larger */ #endif typedef union { --- 68,77 ---- #define BROADCAST_ADDR ((ipaddr_t)0xffffffff) #define LOCALHOST_ADDR ((ipaddr_t)0x7f000001) ! #if PACKETSZ > 512 ! #define MAXPACKET PACKETSZ /* PACKETSZ should be the max udp size (512) */ #else ! #define MAXPACKET 65536 /* but tcp packets can be considerably larger */ #endif typedef union { From jim at rfc1035.com Mon Nov 26 23:41:15 2001 From: jim at rfc1035.com (Jim Reid) Date: Mon, 26 Nov 2001 22:41:15 +0000 Subject: host tool used for RIPE hostcounts In-Reply-To: Your message of "Mon, 26 Nov 2001 15:34:38 GMT." <20011126153438.G43008@isnic.is> Message-ID: <19425.1006814475@gromit.rfc1035.com> >>>>> "Olafur" == Olafur Osvaldsson writes: Olafur> We've been using the host tool from ftp.ripe.net for the Olafur> hostcounts since I don't remember when but noticed some Olafur> strange behaviour this month. Olafur> We got a bunch of these errors: Olafur> *** incomplete resource record for Olafur> ppp129-110.islandssimi.is, offset 8186 ;; response Olafur> truncated to 8192 bytes ;; got answer, 23618 bytes: ;; Olafur> ns_initparse: Message too long *** expand error in Olafur> resource record for islandssimi.is, offset 8186 Olafur> After some diging it looks like this only happens when Olafur> transfering from nameservers with version 9 of bind, wich Olafur> seems to be sending the whole transfer at once instead of Olafur> sending one RR at a time like bind 8 does. Er not quite. The default mode for zone transfers in BIND9 is many-answers. It tries to pack as many RRs into each packet as possible -- not necessarily the whole zone! -- instead of just one RR at a time. Who maintains that host program now? ISTR Brad Knowles mumbling something about taking on that responsibility. From Daniel.Karrenberg at ripe.net Tue Nov 27 09:00:46 2001 From: Daniel.Karrenberg at ripe.net (Daniel Karrenberg) Date: Tue, 27 Nov 2001 09:00:46 +0100 Subject: host tool used for RIPE hostcounts In-Reply-To: <19425.1006814475@gromit.rfc1035.com> References: Message-ID: <4.3.2.7.2.20011127090004.03e803e0@localhost.ripe.net> At 11:41 PM 26.11.01, Jim Reid wrote: >Who maintains that host program now? ISTR Brad Knowles mumbling >something about taking on that responsibility. Err, this has fallen into a very big crack here in Amsterdam. Let me sort this out. Daniel