x
git-svn-id: svn://svn.h5l.se/heimdal/trunk/heimdal@4045 ec53bebd-3082-4978-b11e-865c3cabbd6b
This commit is contained in:
		| @@ -14,11 +14,13 @@ | ||||
| .ad l | ||||
| .in 0 | ||||
| .ta \n(.luR | ||||
| .nf | ||||
| Network Working Group	Assar Westerlund | ||||
| <draft-ietf-cat-krb5-tcp.txt>	SICS | ||||
| Internet-Draft	Johan Danielsson | ||||
| October, 1997	PDC, KTH | ||||
| Expire in six months	 | ||||
| November, 1997	PDC, KTH | ||||
| Expire in six months | ||||
| .fi | ||||
|  | ||||
| .ce | ||||
| Kerberos over TCP | ||||
| @@ -58,25 +60,24 @@ protocol. | ||||
| .ti 0 | ||||
| Specification | ||||
|  | ||||
| A Kerberos server MAY accepts requests on port 88 (decimal). | ||||
| This draft specifies an extension to section 8.2.1 of RFC1510.  | ||||
|  | ||||
| A Kerberos server MAY accept requests on TCP port 88 (decimal). | ||||
|  | ||||
| The data sent from the client to the KDC should consist of 4 bytes | ||||
| containting the length Kerberos request in network byte order and then | ||||
| the request (AS-REQ or TGS-REQ) itself.  The reply from the KDC should | ||||
| consist of the length of the reply packet in network byte order | ||||
| followed by the packet itself (AS-REP, TGS-REP, or KRB-ERROR). | ||||
| containing the length, in network byte order, of the Kerberos request, | ||||
| followed by the request (AS-REQ or TGS-REQ) itself.  The reply from | ||||
| the KDC should consist of the length of the reply packet (4 bytes, | ||||
| network byte order) followed by the packet itself (AS-REP, TGS-REP, or | ||||
| KRB-ERROR). | ||||
|  | ||||
| C->S: Open connection to port 88 at the server | ||||
| .br | ||||
| C->S: length of request [as 4 bytes in network byte order ] | ||||
| .br | ||||
| .nf | ||||
| C->S: Open connection to TCP port 88 at the server | ||||
| C->S: length of request | ||||
| C->S: AS-REQ or TGS-REQ | ||||
| .br | ||||
| S->C: length of reply [as 4 bytes in network byte order ] | ||||
| .br | ||||
| S->C: length of reply | ||||
| S->C: AS-REP, TGS-REP, or KRB-ERROR | ||||
| .br | ||||
|     : connection is closed | ||||
| .fi | ||||
|  | ||||
| .ti 0 | ||||
| Discussion | ||||
| @@ -90,12 +91,11 @@ than UDP. | ||||
| In theory, there's no reason for having explicit length fields, that | ||||
| information is already encoded in the ASN1 encoding of the Kerberos | ||||
| packets.  But having explicit lengths makes it unnecessary to have to | ||||
| decode the ASN1 encoding just to know how much data has to be read. | ||||
| decode the ASN.1 encoding just to know how much data has to be read. | ||||
|  | ||||
| Another way of signaling the end of the request of the reply would be | ||||
| to do a half-close after the request and a full-close after the | ||||
| reply.  This does not work well with all kind of firewalls so the | ||||
| simpler and more robust solution of explicit length fields. | ||||
| to do a half-close after the request and a full-close after the reply. | ||||
| This does not work well with all kinds of firewalls. | ||||
|  | ||||
| .ti 0 | ||||
| Security considerations | ||||
|   | ||||
		Reference in New Issue
	
	Block a user
	 Johan Danielsson
					Johan Danielsson