[230] in Hesiod

home help back first fref pref prev next nref lref last post

Re: huge printcap entries

daemon@ATHENA.MIT.EDU (Randall S. Winchester)
Fri Jul 29 14:54:35 1994

From: "Randall S. Winchester" <rsw@Glue.umd.edu>
Date: Fri, 29 Jul 1994 14:25:35 -0400
To: greg@duke.CS.UNLV.EDU
Cc: bind-workers@vix.com, db74+@andrew.cmu.edu, hesiod@MIT.EDU,
        walsh@athena.tay.dec.com

I have always had to change arpa/nameserv.h like this;
#define PACKETSZ        8192            /* maximum packet size */
/* #define PACKETSZ     512             /* maximum packet size */

The hesiod source uses this PACKETSZ. 

I have always compiled named/libresolv.a with PACKETSZ = 8192.
My (lack  of) understanding was that the 512 number has been carried along 
historicaly, and that the relevent RFC's relating to Text based records 
allow for this.

I have some HS records that are kind of long ( not a printcap mind you)
and this is the only way I have ever been able to get the data across.
I have run this way with BIND 4.8.3, 4.9.2, and now 4.9.3BETA9p1 and this change
has never seemed to cause any problems.

Randall

> From greg@duke.CS.UNLV.EDU Fri Jul 29 13:33:54 1994
> Cc: Dan Walsh <walsh@athena.tay.dec.com>,
>         "Derrick J. Brashear" <db74+@andrew.cmu.edu>, hesiod@MIT.EDU,
>         bind-workers@vix.com
> Subject: Re: huge printcap entries 
> From: Greg Wohletz <greg@duke.CS.UNLV.EDU>
> 
> >>There are two places where you need to change the limit.  
> >>
> >>First you need to change hesiod to use a tcp connection. 
> >>
> >
> >I've always thought that hesiod should use a VC.  I'll give this a try.
> 
> Even with these changes there is still a problem.  Although the hesiod
> program will use a VC to connect to the local server, the local server
> won't use a VC to contact the authoritative server, so you are right back
> where you started from.  
> 
> Perhaps someone could comment on how the current verion of BIND handles large
> answers.  From the debugging output it would appear that the largest packet
> it is willing to send is 504 bytes despite using the 8k buffer size.
> Perhaps the forwarder should use a VC if it receives a truncated answer on
> its 1st try.
> 
> 						--Greg
> 

home help back first fref pref prev next nref lref last post