Network Working Group D. New
Request for Comments: 3195 M. Rose
Category: Standards Track Dover Beach Consulting, Inc.
November 2001
Reliable Delivery for syslog
Status of this Memo
This document specifies an Internet standards track protocol for the
Internet community, and requests discussion and suggestions for
improvements. Please refer to the current edition of the "Internet
Official Protocol Standards" (STD 1) for the standardization state
and status of this protocol. Distribution of this memo is unlimited.
Copyright Notice
Copyright (C) The Internet Society (2001). All Rights Reserved.
Abstract
The BSD Syslog Protocol describes a number of service options related
to propagating event messages. This memo describes two mappings of
the syslog protocol to TCP connections, both useful for reliable
delivery of event messages. The first provides a trivial mapping
maximizing backward compatibility. The second provides a more
complete mapping. Both provide a degree of robustness and security
in message delivery that is unavailable to the usual UDP-based syslog
protocol, by providing encryption and authentication over a
connection-oriented protocol.
New & Rose Standards Track [Page 1]
RFC 3195 Reliable Delivery for syslog November 2001
Table of Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 3
2. The Model . . . . . . . . . . . . . . . . . . . . . . . . . 4
3. The RAW Profile . . . . . . . . . . . . . . . . . . . . . . 7
3.1 RAW Profile Overview . . . . . . . . . . . . . . . . . . . . 7
3.2 RAW Profile Identification and Initialization . . . . . . . 9
3.3 RAW Profile Message Syntax . . . . . . . . . . . . . . . . . 10
3.4 RAW Profile Message Semantics . . . . . . . . . . . . . . . 10
4. The COOKED Profile . . . . . . . . . . . . . . . . . . . . . 11
4.1 COOKED Profile Overview . . . . . . . . . . . . . . . . . . 11
4.2 COOKED Profile Identification and Initialization . . . . . . 11
4.3 COOKED Profile Message Syntax . . . . . . . . . . . . . . . 11
4.4 COOKED Profile Message Semantics . . . . . . . . . . . . . . 12
4.4.1 The IAM Element . . . . . . . . . . . . . . . . . . . . . . 12
4.4.2 The ENTRY Element . . . . . . . . . . . . . . . . . . . . . 14
4.4.3 The PATH Element . . . . . . . . . . . . . . . . . . . . . . 19
5. Additional Provisioning . . . . . . . . . . . . . . . . . . 25
5.1 Message Authenticity . . . . . . . . . . . . . . . . . . . . 25
5.2 Message Replay . . . . . . . . . . . . . . . . . . . . . . . 25
5.3 Message Integrity . . . . . . . . . . . . . . . . . . . . . 25
5.4 Message Observation . . . . . . . . . . . . . . . . . . . . 26
5.5 Summary of Recommended Practices . . . . . . . . . . . . . . 26
6. Initial Registrations . . . . . . . . . . . . . . . . . . . 27
6.1 Registration: The RAW Profile . . . . . . . . . . . . . . . 27
6.2 Registration: The COOKED Profile . . . . . . . . . . . . . . 27
7. The syslog DTD . . . . . . . . . . . . . . . . . . . . . . . 28
8. Reply Codes . . . . . . . . . . . . . . . . . . . . . . . . 32
9. IANA Considerations . . . . . . . . . . . . . . . . . . . . 33
9.1 Registration: BEEP Profiles . . . . . . . . . . . . . . . . 33
9.2 Registration: The System (Well-Known) TCP port number for
syslog-conn . . . . . . . . . . . . . . . . . . . . . . . 33
10. Security Considerations . . . . . . . . . . . . . . . . . . 34
11. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 34
12. References . . . . . . . . . . . . . . . . . . . . . . . . . 34
Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . . 35
Full Copyright Statement . . . . . . . . . . . . . . . . . . . . . 36
New & Rose Standards Track [Page 2]
RFC 3195 Reliable Delivery for syslog November 2001
1. Introduction
The syslog protocol [1] presents a spectrum of service options for
provisioning an event-based logging service over a network. Each
option has associated benefits and costs. Accordingly, the choice as
to what combination of options is provisioned is both an engineering
and administrative decision. This memo describes how to realize the
syslog protocol when reliable delivery is selected as a required
service. It is beyond the scope of this memo to argue for, or
against, the use of reliable delivery for the syslog protocol.
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
"SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this
document are to be interpreted as described in RFC 2119 [2].
New & Rose Standards Track [Page 3]
RFC 3195 Reliable Delivery for syslog November 2001
2. The Model
The syslog service supports three roles of operation: device, relay,
and collector.
Devices and collectors act as sources and sinks, respectively, of
syslog entries. In the simplest case, only a device and collector
are present. E.g.,
+--------+ +-----------+
| Device | -----> | Collector |
+--------+ +-----------+
The relationship between devices and collectors is potentially many-
to-many. I.e., a device might communicate with many collectors;
similarly, a collector might communicate with many devices.
A relay operates in both modes, accepting syslog entries from devices
and other relays and forwarding those entries to collectors and other
relays.
For example,
+--------+ +-------+ +-------+ +-----------+
| Device | ---> | Relay | -...-> | Relay | ---> | Collector |
+--------+ +-------+ +-------+ +-----------+
As shown, more than one relay may be present between any particular
device and collector.
A relay may be necessary for administrative reasons. For example, a
relay might run as an application proxy on a firewall. Also, there
might be one relay per company department, which authenticates all
the devices in the department, and which in turn authenticates itself
to a company-wide collector.
A relay can also serve to filter messages. For example, one relay
may collect the syslog information from an entire web server farm,
summarizing hit counts for report generation, forwarding "page not
found" messages (indicating a possible broken link) to a collector
that presents it to the webmaster, and sending more urgent messages
(such as hardware failure reports) to a collector that gateways them
to a pager. A relay may also be used to convert formats from a
device's output to a collector's input.
It should be noted that a role of device, relay, or collector is
relevant only to a particular BEEP channel (q.v., below). A single
server can serve as a device, a relay, and a collector, all at once,
New & Rose Standards Track [Page 4]
RFC 3195 Reliable Delivery for syslog November 2001
if so configured. It can even serve as a relay and a collector to
the same device at the same time using different BEEP channels over
the same connection-oriented session; this might be useful to collect
status yet relay urgent error messages.
To provide reliable delivery when realizing the syslog protocol, this
memo defines two BEEP profiles. BEEP [3] is a generic application
protocol framework for connection-oriented, asynchronous
interactions. Within BEEP, features such as authentication, privacy,
and reliability through retransmission are provided. There are two
profiles defined in this memo:
o The RAW profile is designed to provide a high-performance, low-
impact footprint, using essentially the same format as the
existing UDP-based syslog service.
o The COOKED profile is designed to provide a structured entry
format, in which individual entries are acknowledged (either
positively or negatively).
Note that both profiles run over BEEP. BEEP defines "transport
mappings," specifying how BEEP messages are carried over the
underlying transport technologies. At the time of this writing, only
one such transport is defined, in [4], which specifies BEEP over TCP.
All transport mappings are required to support enough reliability and
sequencing to allow all BEEP messages on a given channel to be
delivered reliably and in order. Hence, both the RAW and COOKED
profile provide reliable delivery of their messages.
The choice of profile is independent of the operational roles
discussed above.
For example, in
+--------+ +-------+ +-----------+
| Device | -----> | Relay | -----> | Collector |
+--------+ +-------+ +-----------+
the device-to-relay link could be configured to use the RAW profile,
while the relay-to-collector link could be configured to use the
COOKED profile. (For example, the relay may be parsing the RAW
syslog messages from the device, knowing the details of their
formats, before passing them to a more generic collector.) Indeed,
the same device may use different profiles, depending on the
collector to which it is sending entries.
New & Rose Standards Track [Page 5]
RFC 3195 Reliable Delivery for syslog November 2001
Devices and relays MAY discover relays and collectors via the DNS SRV
algorithm [5]. If so configured, the service used is "syslog" and
the protocol used is "tcp". This allows for central administration
of addressing, fallback for failed relays and collectors, and static
load balancing. Security policies and hardware configurations may be
such that device configuration is more secure than the DNS server.
Hardware devices may be of such limited resources that DNS SRV access
is inappropriate. Firewalls and other restrictive routing mechanisms
may need to be dealt with before a reliable syslog connection can be
established. In these cases, DNS might not be the most appropriate
configuration mechanism.
New & Rose Standards Track [Page 6]
RFC 3195 Reliable Delivery for syslog November 2001
3. The RAW Profile
3.1 RAW Profile Overview
The RAW profile is designed for minimal implementation effort, high
efficiency, and backwards compatibility. It is appropriate
especially in cases where legacy syslog processing will be applied.
It should be noted that even though the RAW profile uses the same
format for message payloads as the UDP version of syslog uses,
delivery is reliable. The RAW syslog profile is a profile of BEEP
[3], and BEEP guarantees ordered reliable delivery of messages within
each individual channel.
When the profile is started, no piggyback data is supplied. All BEEP
messages in the RAW profile are specified as having a MIME Content-
Type [6] of application/octet-stream. Once the channel is open, the
listener (not the initiator) sends a MSG message indicating it is
ready to act as a syslog sink. (Refer to [3]'s Section 2.1 for a
discussion of roles that a BEEP peer may perform, including
definitions of the terms "listener", "initiator", "client", and
"server".)
The initiator uses ANS replies to supply one or more syslog entries
in the current UDP format, as specified in [1]'s Section 3. When the
initiator has no more entries to send, it finishes with a NUL reply
and closes the channel.
An example might appear as follows:
L:
I:
L: RPY 0 0 . 0 201
L: Content-type: application/beep+xml
L:
L:
L:
L: END
I: RPY 0 0 . 0 52
I: Content-type: application/beep+xml
I:
I:
I:
I: END
L: RPY 0 1 . 201 100
L: Content-type: application/beep+xml
L:
L:
I:
L:
New & Rose Standards Track [Page 8]
RFC 3195 Reliable Delivery for syslog November 2001
Here we see a BEEP session established, followed by the use of the
RAW profile. The initiator is a device, while the listener is a
collector. The initiator opens the channel, but the listener sends
the first MSG. This allows the initiator to send any number of ANS
replies carrying syslog event messages. The initiator sends a NUL
reply to indicate it is finished. Upon receiving the NUL, the
listener closes the RAW channel. The initiator has the choice of
closing the entire BEEP session or opening a new syslog channel (RAW
or COOKED) for more transfers. In this example, the initiator
chooses to close the entire BEEP session.
The overhead for one ANS frame is about thirty octets, once the
initial handshakes have been exchanged. If this overhead is too
high, then messages are likely being generated at a high rate. In
this case, multiple syslog messages can be aggregated into a single
ANS frame, each separated by a CRLF sequence from the preceding. The
final message still MUST NOT end with a CRLF.
For example,
L: MSG 1 0 . 0 50
L:
L: Central Services. This has not been a recording.
L: END
I: ANS 1 0 . 0 119 0
I:
I: <29>Oct 27 13:21:08 ductwork imxpd[141]: Heating emergency.
I: <29>Oct 27 13:21:09 ductwork imxpd[141]: Contact Tuttle.END
I: NUL 1 0 . 119 0
I: END
3.2 RAW Profile Identification and Initialization
The RAW syslog profile is identified as
http://xml.resource.org/profiles/syslog/RAW
in the BEEP "profile" element during channel creation.
No data is piggybacked during channel creation.
New & Rose Standards Track [Page 9]
RFC 3195 Reliable Delivery for syslog November 2001
3.3 RAW Profile Message Syntax
All BEEP messages in this profile have a MIME content-type of
application/octet-stream. The listener's first BEEP message is
ignored and indeed may be empty except for headers; hence, any syntax
is acceptable.
The ANS replies the initiator sends in response MUST be formatted
according to Section 4 of [1]. In particular, If the receiver is
acting as a relay, then it MUST follow the rules as laid out in
Section 4.2.2 of [1].
If multiple syslog messages are included in a single ANS reply, each
is separated from the preceding with a CRLF. There is no ending
delimiter, but each syslog event message body length MUST be 1024
bytes or less, excluding BEEP framing overhead. Note that there MUST
NOT be a CRLF between the text of the final syslog event message and
the "END" marking the trailer of the BEEP frame.
3.4 RAW Profile Message Semantics
The listener's opening BEEP MSG message has no semantics. (It is a
good place to put in an identifying greeting.) The initiator's ANS
replies MUST specify a facility, severity, and textual message, as
described in [1].
New & Rose Standards Track [Page 10]
RFC 3195 Reliable Delivery for syslog November 2001
4. The COOKED Profile
4.1 COOKED Profile Overview
The COOKED profile is designed for new implementations of syslog
protocol handlers. It provides a much finer grain of information
tagging, allowing a better degree of automation in processing.
Naturally, it includes more overhead as well in support of this.
The COOKED profile supports three elements of interest:
o The "iam" element identifies the sender to the receiver, allowing
each peer to name itself for the other, and specifying the roles
(device, relay, or collector) each is taking on.
o The "entry" element provides a parsed version of the syslog entry,
with the various fields of interest broken out.
o The "path" element identifies a list of relays through which a
tagged collection of "entry" elements has passed, along with a set
of flags indicating what assurances of security have been in
effect throughout its delivery.
4.2 COOKED Profile Identification and Initialization
The COOKED syslog profile is identified as
http://xml.resource.org/profiles/syslog/COOKED
in the BEEP "profile" element during channel creation.
During channel creation, the corresponding "profile" element in the
BEEP "start" element may contain an "iam" element. If channel
creation is successful, then before sending the corresponding reply,
the BEEP peer processes the "iam" element and includes the resulting
response in the reply. This response will be an "ok" element or an
"error" element. The choice of which element is returned is
dependent on local provisioning of the recipient. Including an "iam"
in the initial "start" element has exactly the same semantics as
passing it as the first MSG message on the channel.
4.3 COOKED Profile Message Syntax
All BEEP messages in this profile have a MIME Content-Type [6] of
application/beep+xml. The syntax of the individual elements is
specified in Section 7.
New & Rose Standards Track [Page 11]
RFC 3195 Reliable Delivery for syslog November 2001
4.4 COOKED Profile Message Semantics
Initiators issue two elements: "iam" and "entry", each using a "MSG"
message. The listener issues "ok" in "RPY" messages and "error" in
"ERR" messages. (See [3]'s Section 2.3.1 for the definitions of the
"error" and "ok" elements.)
4.4.1 The IAM Element
The "iam" element serves to identify a device, relay, or collector at
one end of the BEEP channel to the device, relay, or collector at the
other end of the channel. The "iam" element includes the type of
peer (device, relay, or collector), the fully qualified domain name
of the peer, and an IP address of the peer. (The IP address chosen
SHOULD be the IP address associated with the underlying transport
protocol carrying the channel.) The character data of the element is
free-form human-readable text. It may be used to further identify
the peer, such as by describing the physical location of the machine.
An "iam" element may be sent by the initiator of the channel at any
time. The listener responds to an "iam" element with an "ok"
(indicating acceptance), or an "error" (indicating rejection). The
identity and role in effect is specified by the most recent "iam"
answered with an "ok".
An "iam" could be rejected (with an "error" element) by the listener
if the privacy or authentication that has been negotiated is
inadequate or if the authenticated user does not have authorization
to serve in the specified role. It is expected that most
installations will require an "iam" from the peer before accepting
any "entry" messages.
New & Rose Standards Track [Page 12]
RFC 3195 Reliable Delivery for syslog November 2001
For example, a successful creation might look like this:
I: MSG 0 10 . 1832 259
I: Content-type: application/beep+xml
I:
I:
I:
I: ]]>
I:
I:
L: END
L: RPY 0 10 . 704 138
L: Content-type: application/beep+xml
L:
L:
L: ]]>
L:
L: END
A creation with an embedded "iam" that fails might look like this:
C: MSG 0 12 . 1832 259
C: Content-type: application/beep+xml
C:
C:
C:
C: ]]>
C:
C:
C: END
S: RPY 0 12 . 704 241
S: Content-type: application/beep+xml
S:
S:
S: User 'buttle.example.com' not allowed
S: to "iam" for 'tuttle.example.com' ]]>
S:
S: END
In this case, the error code indicates that the user
"buttle.example.com" has logged in via some SASL profile, but the
syslog COOKED profile implementation is claiming to be
"tuttle.example.com", a mismatch that the server is disallowing.
New & Rose Standards Track [Page 13]
RFC 3195 Reliable Delivery for syslog November 2001
4.4.2 The ENTRY Element
The "entry" element carries the details of a single syslog entry. The
attributes of an "entry" element include "facility", "severity",
"timestamp", "hostname", and "tag". "Facility" and "severity" have
the semantics defined in [1]'s 4.1. The other attributes have the
semantics as in Sections 4.2.1 and 4.2.3 of [1]. An "entry" element
can also contain a "pathID" attribute, described below.
If the client is a relay, the "entry" SHOULD also contain the
attributes "deviceFQDN" and "deviceIP", specifying the FQDN and IP
address of the device that originally created the entry. These
attributes may be added by either the relay or the originating
device. If possible, the device SHOULD add these entries, referring
to the interface most closely associated with the syslog entry.
Before a relay forwards an entry from a device that does not carry
these attributes, it SHOULD add them based on the "iam" element it
has received from the device, or based on the underlying transport
connection address. A relay MUST NOT add these fields if they are
missing and an "iam" element on the channel has indicated that
messages are coming from another relay.
The "pathID" attribute indicates the path over which this entry has
travelled, from device through relays to the final collector.
Syntactically, its value is a string of digits that must match the
"pathID" attribute of a "path" element sent earlier over the current
channel. Semantically, it indicates that the list of relays and
flags indicated in that earlier "path" element apply to this "entry"
element.
The character data for the element is the unstructured syslog event
message being logged. If the original device delivers the message
for the first time via the COOKED profile, it may have any structure
inside the CDATA. However, for maximum compatibility, the device
SHOULD format the CDATA of the message in accordance with Sections
4.2.1 through 4.2.3 of [1].
New & Rose Standards Track [Page 14]
RFC 3195 Reliable Delivery for syslog November 2001
In the message is being relayed, "tag" SHOULD be those of the
original device generating the entry (unless the device cannot supply
a tag). The "timestamp" SHOULD be that of the original entry
generation time, rather than the time the entry was passed outward
from the relay. The "hostname" SHOULD be the host name or IP address
by which the device knows itself; this MUST follow the rules
established in Sections 4.2.1 through 4.2.3 of [1]. The original
contents of the syslog message MUST be preserved in the CDATA of the
"entry" element; this includes preservation of exact content during
translation from the UDP or RAW formats. In particular, the
timestamps MUST NOT be rewritten in the CDATA of the "entry" element,
the tag MUST NOT be removed from the CDATA even if presented in the
"entry" attributes as well, and so on.
To be consistent with the spirit of [1], a relay receiving a message
that does not contain a valid priority, timestamp or hostname will
follow the same general rules as described in section 4.2.2 of [1]
while including the exact contents of the received syslog packet as
the CDATA. The values of the facility and severity will be construed
to be 8 and 6 respectively and will be placed into the appropriate
attributes of the "entry" element. The hostname will be the name of
the device as it is known to the relay and will also be inserted into
the "entry" element's attributes. The timestamp would be set to the
received time, inserted only into the attributes of the "entry"
element. As an example, consider this message received on UDP port
514 and interpreted as a traditional syslog message, assuming the
underlying IP source address is that of the "pipeworks" machine:
<.....eeeek!
To be relayed, it must be modified as follows:
C: MSG 1 0 . 2079 156
C: Content-Type: application/beep+xml
C:
C: <.....eeeek!
C: END
S: RPY 1 0 . 933 45
S: Content-Type: application/beep+xml
S:
S: <166> 1990 Oct 22 01:00:00 bomb tick[0]: BOOM!
C: END
S: RPY 1 0 . 978 45
S: Content-Type: application/beep+xml
S:
S: <166> Oct 22 01:00:00 bomb tick[0]: BOOM!
C: END
S: RPY 1 0 . 1023 45
S: Content-Type: application/beep+xml
S:
S:
C: No 27B/6 available
C: END
S: RPY 1 0 . 1068 45
S: Content-Type: application/beep+xml
S:
S:
C: Replacement device found in nostril.
C:
C: END
S: ERR 1 2 . 1113 111
S: Content-Type: application/beep+xml
S:
S: Not allowed to relay for
S: jack.example.net
S: END
Here, the client attempts to relay an entry on behalf of
jack.example.com, but the entry is refused by the collector for
administrative reasons. This may occur, for example, if
lowry.example.com is in a different department than jack.example.com.
New & Rose Standards Track [Page 18]
RFC 3195 Reliable Delivery for syslog November 2001
4.4.3 The PATH Element
The "path" element serves to describe a list of the relays through
which that element has passed, along with a set of flags that
indicate the properties that all links from the device to the relay
have shared in common. Each "path" element contains either another
"path" element or is empty. An empty "path" element identifies a
device, while a "path" element with a nested "path" element
identifies a relay. Each "path" element names a FQDN and IP address
of the interface that sent the element. Each "path" element also
names a FQDN and IP address for the interface that received the
element. Each "path" element also carries a "linkprops" attribute,
specifying the properties of the link it describes.
Each "path" element has a "pathID" attribute which must be unique for
all "path" elements sent on this channel since its inception.
Syntactically, the "pathID" attribute is a string of digits.
Semantically, it serves to identify one "path" element out of many,
and it serves to link a "path" element with one or more "entry"
elements. Any "pathID" attribute is unrelated to any "pathID"
attribute in nested "path" elements or on other channels.
Each "path" element has a "fromFQDN" attribute and an "fromIP"
attribute. The "fromFQDN" attribute SHOULD be the fully qualified
domain name of the interface over which the "path" element was sent.
(The "fromFQDN" can be omitted if that interface has no DNS entry.)
Similarly, the "fromIP" attribute MUST be the IP address of the
interface over which the "path" element was sent.
Each "path" element has a "toFQDN" attribute and an "toIP" attribute.
The "toFQDN" attribute SHOULD be the fully qualified domain name of
the interface over which the "path" element was received. (The
"toFQDN" can be omitted if that interface has no DNS entry.)
Similarly, the "toIP" attribute MUST be the IP address of the
interface over which the "path" element was received.
Finally, each "path" element carries a "linkprops" attribute. This
is syntactically a string of individual characters, each indicating
one property of the channel over which this "path" element is being
carried. Note that outer "path" elements may have stronger
guarantees than inner "path" elements; care should be taken in the
interpretation of flags. The semantics of each possible character in
this string are as follows:
New & Rose Standards Track [Page 19]
RFC 3195 Reliable Delivery for syslog November 2001
o: When present, "o" (lower-case letter "o") indicates that weak
privacy has been negotiated over this link, weakly protecting from
observation the content of entries associated with this "path"
element. (Weak privacy is encryption with less than 80 bits of
key.)
O: When present, "O" (upper-case letter "O") indicates that strong
privacy has been negotiated over this link, strongly protecting
from observation the content of entries associated with this
"path" element. (Strong privacy is encryption with 80 bits or
more of key, or a transfer mechanism that is otherwise impossible
to eavesdrop upon.)
U: When present, "U" indicates that a valid user has been
authenticated (via SASL or TLS) and an "iam" element has been
accepted.
A: When present, "A" indicates that this link has been protected by
an authentication layer, authenticating the source of every
"entry" associated with this path.
R: When present, "R" indicates that this link has been protected
against message replay.
I: When present, "I" indicates that this link has been protected
against modifications of messages in passing. ("I" stands for
message Integrity.)
L: When present, "L" indicates that this link has been protected
against loss of messages. That is, this is a reliable delivery
link.
D: When present, "D" indicates that the "from" side of this link is a
device. If this is not present on the innermost "path" element,
"entry" elements associated with this path have not been carried
by the COOKED profile for their entire lifetime.
Upon receiving a "path" element, the peer MUST perform the following
checks:
o The "fromFQDN" and "fromIP" must match the underlying transport
connection.
o The flags in the "linkprops" attribute must match the attributes
of the session.
o The "toFQDN" and "toIP" must match the underlying transport
connection.
New & Rose Standards Track [Page 20]
RFC 3195 Reliable Delivery for syslog November 2001
o The "pathID" attribute must be unique with respect to all other
"path" elements received on this channel.
If all these checks pass, the "path" element is accepted with an "ok"
element. Otherwise, an "error" element is generated with an
appropriate code. In addition, if any of the nested "path" elements
refer to the machine receiving the element, it may indicate a routing
loop in the configuration for the so-identified path, and appropriate
measures should be taken.
If the peer receiving an "entry" element is receiving it directly
from a device via either syslog-conn profile, and the device has not
generated a "path" element, the receiver may itself generate an
appropriate "path" element, either to be recorded in the logs (if
this peer is a collector) or passed to the next peer (if this peer is
a relay). If a peer receives a syslog message via UDP, it may
optionally generate an appropriate "peer" element based on any
cryptographic information provided in the message itself.
When a peer receives a "path" element, it remembers it for future
use. A collector will store it in the log for later reference. A
relay will remember it. When an "entry" arrives referencing the
received "path" element, and that entry needs to be forwarded to
another relay or collector, and no appropriate "path" element has
already been generated, an appropriate "path" element is generated
and sent over the outbound channel before the entry is forwarded. An
appropriate "path" element is created by taking the received "path"
element, wrapping it in a new "path" element with the appropriate
attributes, and assigning it a new "pathID" attribute. When future
"entry" elements arrive with the same incoming "pathID" attribute,
and they need to be forwarded to a channel over which an appropriate
"pathID" attribute has already been sent, only the "pathID" attribute
of the "entry" element needs to be rewritten to refer to the "path"
element on the outgoing channel.
It should be noted that the majority of the complexity in managing
"path" elements arises only in relays. In particular, devices never
need to generate "path" elements and collectors need only verify
them, log them, and possibly use them in displays and reports.
Collectors do not need to generate "path" elements or rewrite "entry"
elements. Hence, only in complex configurations (where they are most
useful) do complex "path" configurations occur.
New & Rose Standards Track [Page 21]
RFC 3195 Reliable Delivery for syslog November 2001
For example, here is a path element sent from
lowry.records.example.com to kurtzman.records.example.com. It
indicates that entries from lowry to kurtzman tagged with
pathID='173' originated from screen.lowry.records.example.com. It
indicates that screen.lowry.records.example.com is believed by
lowry.records.example.com to be the originating device, and that
entries over this path are delivered without loss and without
modification, although messages might be replayed or observed. The
link between lowry and kurtzman, however, avoids replay attacks, lost
messages, and modifications to messages. While
screen.lowry.records.example.com has not authenticated itself to
lowry.records.example.com, lowry claims to have authenticated itself
to kurtzman.
C: MSG 2 1 . 3121 426
C: Content-type: application/beep+xml
C:
C:
C:
C:
C:
C: END
S: ERR 2 1 . 1224 114
S: Content-type: application/beep+xml
S:
S: linkprops includes 'U'
S: but no 'iam' received
S: END
However, kurtzman.records.example.com rejects the "path" element,
since the "linkprops" attribute claims that lowry has authenticated
itself, but kurtzman disagrees, not having received an "iam" element.
New & Rose Standards Track [Page 22]
RFC 3195 Reliable Delivery for syslog November 2001
In a second example, this "path" element informs
collector.example.com that the records department's firewall will be
forwarding "entry" elements with a "pathID" attribute whose value is
"17". These "entry" elements will be coming in on the "10.0.0.2"
interface of the firewall, to be forwarded out the "134.130.74.56"
interface of the firewall. The final hop has all possible
guarantees, although the entries transferred within the records
department (behind the firewall) may have been observed in passing.
C: MSG 2 2 . 3547 813
C: Content-type: application/beep+xml
C:
C:
C:
C:
C:
C:
C: END
S: RPY 2 2 . 1338 45
S: Content-type: application/beep+xml
S:
S:
C: Job paused - Boss watching.
C:
C: END
S: RPY 2 3 . 1383 45
S: Content-Type: application/beep+xml
S:
S:
C: Job paused - Boss watching.
C:
C: END
S: RPY 7 9 . 338 45
S: Content-Type: application/beep+xml
S:
S:
RFC, FYI, BCP