RFC 7014
Flow Selection Techniques
Flow Selection Techniques 💖
RFC 7014 🎀 is a Standards Track document (category: Standards Track, status: Internet Standards Track), and its ISSN is 2070-1721. It came out in September 2013. Nothing in its header says it obsoletes or updates any other RFC. The authors are S. D'Antonio (Univ. of Napoli "Parthenope"), T. Zseby (CAIDA/FhG FOKUS), C. Henke (Tektronix Communications Berlin), and L. Peluso (University of Napoli). This is giving serious research-team energy 🧚✨
Abstract
okay so the Intermediate Flow Selection Process is basically the process of picking a subset of Flows out of all the Flows you've observed 🦋 It can be located at an IP Flow Information Export (IPFIX) Exporter or Collector, or inside an IPFIX Mediator. The whole point is that it cuts down the effort of post-processing Flow data and transferring Flow Records 💅 This doc explains why you'd want to use the Intermediate Flow Selection Process, walks through the actual selection techniques, gives an information model for configuring them, and talks about what information about the process should be exported. iconic, honestly 👑
Status of This Memo
This is an Internet Standards Track document 🌸
This document is a product of the Internet Engineering Task Force (IETF). It represents the consensus of the IETF community. It went through public review and got approved for publication by the Internet Engineering Steering Group (IESG). If you want more info on Internet Standards, check out Section 2 of RFC 5741 🫶
For the current status of this document, any errata, and how to send feedback, go to http://www.rfc-editor.org/info/rfc7014 ✨
Copyright Notice
Copyright (c) 2013 IETF Trust and the persons identified as the document authors. All rights reserved. 💜
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (http://trustee.ietf.org/license-info) as they stand on the date this document was published. Please read these documents carefully, because they explain your rights and restrictions regarding this document 🙈 Any Code Components pulled out of this document must include the Simplified BSD License text described in Section 4.e of the Trust Legal Provisions, and they come with no warranty, as described in the Simplified BSD License.
This document might contain material from IETF Documents or IETF Contributions that were published or made publicly available before November 10, 2008. The person(s) controlling the copyright in some of that material might not have granted the IETF Trust the right to allow modifications outside the IETF Standards Process. Without getting an adequate license from whoever controls the copyright in those materials, this document can't be modified outside the IETF Standards Process, and derivative works of it can't be created outside the IETF Standards Process either. The exceptions are formatting it for publication as an RFC or translating it into languages other than English 🍓
Table of Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.1. Requirements Language . . . . . . . . . . . . . . . . . . 4
2. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . 4
3. Difference between Intermediate Flow Selection Process and
Packet Selection . . . . . . . . . . . . . . . . . . . . . . . 7
4. Difference between Intermediate Flow Selection Process and
Intermediate Selection Process . . . . . . . . . . . . . . . . 9
5. Intermediate Flow Selection Process within the IPFIX
Architecture . . . . . . . . . . . . . . . . . . . . . . . . . 9
5.1. Intermediate Flow Selection Process in the Metering
Process . . . . . . . . . . . . . . . . . . . . . . . . . 11
5.2. Intermediate Flow Selection Process in the Exporting
Process . . . . . . . . . . . . . . . . . . . . . . . . . 11
5.3. Intermediate Flow Selection Process as a Function of
the IPFIX Mediator . . . . . . . . . . . . . . . . . . . . 11
6. Intermediate Flow Selection Process Techniques . . . . . . . . 12
6.1. Flow Filtering . . . . . . . . . . . . . . . . . . . . . . 12
6.1.1. Property Match Filtering . . . . . . . . . . . . . . . 12
6.1.2. Hash-Based Flow Filtering . . . . . . . . . . . . . . 13
6.2. Flow Sampling . . . . . . . . . . . . . . . . . . . . . . 13
6.2.1. Systematic Sampling . . . . . . . . . . . . . . . . . 13
6.2.2. Random Sampling . . . . . . . . . . . . . . . . . . . 14
6.3. Flow-State Dependent Intermediate Flow Selection
Process . . . . . . . . . . . . . . . . . . . . . . . . . 14
6.4. Flow-State Dependent Packet Selection . . . . . . . . . . 15
7. Configuration of Intermediate Flow Selection Process
Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . 16
7.1. Intermediate Flow Selection Process Parameters . . . . . . 17
7.2. Description of Flow-State Dependent Packet Selection . . . 19
8. Information Model for Intermediate Flow Selection Process
Configuration and Reporting . . . . . . . . . . . . . . . . . 20
9. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 22
9.1. Registration of Information Elements . . . . . . . . . . . 22
9.1.1. flowSelectorAlgorithm . . . . . . . . . . . . . . . . 22
9.1.2. flowSelectedOctetDeltaCount . . . . . . . . . . . . . 24
9.1.3. flowSelectedPacketDeltaCount . . . . . . . . . . . . . 24
9.1.4. flowSelectedFlowDeltaCount . . . . . . . . . . . . . . 24
9.1.5. selectorIDTotalFlowsObserved . . . . . . . . . . . . . 25
9.1.6. selectorIDTotalFlowsSelected . . . . . . . . . . . . . 25
9.1.7. samplingFlowInterval . . . . . . . . . . . . . . . . . 26
9.1.8. samplingFlowSpacing . . . . . . . . . . . . . . . . . 26
9.1.9. flowSamplingTimeInterval . . . . . . . . . . . . . . . 27
9.1.10. flowSamplingTimeSpacing . . . . . . . . . . . . . . . 27
9.1.11. hashFlowDomain . . . . . . . . . . . . . . . . . . . . 28
9.2. Registration of Object Identifier . . . . . . . . . . . . 28
10. Security and Privacy Considerations . . . . . . . . . . . . . 28
11. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . 30
12. References . . . . . . . . . . . . . . . . . . . . . . . . . . 30
12.1. Normative References . . . . . . . . . . . . . . . . . . . 30
12.2. Informative References . . . . . . . . . . . . . . . . . . 311. Introduction ✨
okay so this document describes Intermediate Flow Selection Process techniques for network traffic measurements 🦋 A Flow is a set of packets that share common properties, as described in [RFC7011]. You can run an Intermediate Flow Selection Process to limit the resources needed for capturing, storing, exporting, and post-processing Flow Records. It can also pick out a particular set of Flows that matter to a specific application. This document categorizes the Intermediate Flow Selection Process techniques and describes the configuration and reporting parameters for each one 💅
This document also covers configuration and reporting parameters for Flow-state dependent packet selection, as described in [RFC5475]. Even though this technique is categorized as packet selection, it's included here because Flow-state dependent packet selection techniques usually aim to reduce the resources used for Flow capturing and Flow processing. Also, [RFC5475] only briefly discussed these techniques, so their configuration and reporting considerations were added to this document 🌈
1.1. Requirements Language
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 [RFC2119]. 🫶
2. Terminology 🎀
Okay bestie, this section is the glossary, and it's literally so important. This document matches the terminology from [RFC7011], [RFC5470], [RFC5475], and [RFC3917] 📚. Like [RFC7011] and [RFC5476], the first letter of each IPFIX specific and Packet Sampling (PSAMP) specific term gets capitalized, and so do the Intermediate Flow Selection Process specific terms defined here. Capital letters are giving main character energy 💅✨
- Packet Classification 🧺 Packet Classification is a process by which packets are mapped to specific Flow Records, based on packet properties or external properties (e.g., interface). The properties (e.g., header information, packet content, Autonomous System (AS) number) make up the Flow Key. If a Flow Record for a specific Flow Key value already exists, the Flow Record is updated; otherwise, a new Flow Record is created.
- Intermediate Flow Selection Process 🦋 An Intermediate Flow Selection Process is an Intermediate Process, as defined in [RFC6183], that takes Flow Records as its input and selects a subset of this set as its output. The Intermediate Flow Selection Process is a more general concept than the Intermediate Selection Process as defined in [RFC6183]. While an Intermediate Selection Process selects Flow Records from a sequence based upon criteria-evaluated Flow Record values and only passes on those Flow Records that match the criteria, an Intermediate Flow Selection Process selects Flow Records using selection criteria applicable to a larger set of Flow characteristics and information. It's the upgraded version, no cap 🔥
- Flow Cache 🗄️ A Flow Cache is the set of Flow Records.
- Flow Selection State 🧠 An Intermediate Flow Selection Process maintains state information for use by the Flow Selector. At a given time, the Flow Selection State may depend on Flows and packets observed at and before that time, as well as other variables. Examples include:
- (i) sequence number of packets and Flow Records;
- (ii) number of selected Flows;
- (iii) number of observed Flows;
- (iv) current Flow Cache occupancy;
- (v) Flow specific counters, lower and upper bounds;
- (vi) Intermediate Flow Selection Process timeout intervals.
- Flow Selector 🎯 A Flow Selector defines the action of an Intermediate Flow Selection Process on a single Flow of its input. The Flow Selector can make use of the following information in order to establish whether or not a Flow has to be selected:
- (i) the content of the Flow Record;
- (ii) any state information related to the Metering Process or Exporting Process;
- (iii) any Flow Selection State that may be maintained by the Intermediate Flow Selection Process.
- Complete Flow 🧩 A Complete Flow consists of all the packets that enter the Intermediate Flow Selection Process within the Flow timeout interval and that belong to the same Flow, per the definition of "Flow" in [RFC5470]. For this definition, only packets that arrive at the Intermediate Flow Selection Process are considered.
- Flow Position 📍 Flow Position is the position of a Flow Record within the Flow Cache.
- Flow Filtering 🚦 Flow Filtering selects flows based on a deterministic function on the Flow Record content, Flow Selection State, external properties (e.g., ingress interface), or external events (e.g., violated Access Control List). If the relevant parts of the Flow Record content can already be observed at the packet level (e.g., Flow Keys from packet header fields), Flow Filtering can be performed at the packet level by Property Match Filtering, as described in [RFC5475].
- Hash-based Flow Filtering 🔐 Hash-based Flow Filtering is a deterministic Flow filter function that selects flows based on a hash function. The hash function is calculated over parts of the Flow Record content or external properties that are called the Hash Domain. If the hash value falls into a predefined Hash Selection Range, the Flow is selected.
- Flow-state Dependent Intermediate Flow Selection Process 🌊 The Flow-state dependent Intermediate Flow Selection Process is a selection function that selects or drops Flows based on the current Flow Selection State. The selection can be either deterministic, random, or non-uniform random.
- Flow-state Dependent Packet Selection 🎁 Flow-state dependent packet selection is a selection function that selects or drops packets based on the current Flow Selection State. The selection can be either deterministic, random, or non-uniform random. Flow-state dependent packet selection can be used to implement a preference for the selection of packets belonging to specific Flows. For example, the selection probability of packets belonging to Flows that are already within the Flow Cache may be higher than for packets that have not been recorded yet. Favoritism, but make it networking 💜
- Flow Sampling 🎲 Flow Sampling selects flows based on Flow Record sequence or arrival times (e.g., entry in Flow Cache, arrival time at Exporter or Mediator). The selection can be systematic (e.g., every n-th Flow) or based on a random function (e.g., select each Flow Record with probability p, or randomly select n out of N Flow Records).
3. Difference between Intermediate Flow Selection Process and Packet Selection ✨
Okay so the Intermediate Flow Selection Process is not the same thing as packet selection from [RFC5475], and this is where people get confused 🙈. Packet selection treats packets as the basic element, and its parent population is every packet observed at an Observation Point. The Intermediate Flow Selection Process flips that. Its basic elements are the Flows, so its parent population is all observed Flows, and it operates on those Flows. The major characteristics of the Intermediate Flow Selection Process are the following:
- The Intermediate Flow Selection Process takes Flows as basic elements. For packet selection, packets are considered as basic elements.
- The Intermediate Flow Selection Process typically takes place after Packet Classification, because the classification rules determine to which Flow a packet belongs. The Intermediate Flow Selection Process can be performed before Packet Classification. In that case, the Intermediate Flow Selection Process is based on the Flow Key (and also on a hash value over the Flow Key) but not on characteristics that are only available after Packet Classification (e.g., Flow size, Flow duration). Packet selection can be applied before and after Packet Classification. As an example, packet selection before Packet Classification can be random packet selection, whereas packet selection after Packet Classification can be Flow-state dependent packet selection (as described in [RFC5475]).
- The Intermediate Flow Selection Process operates on Complete Flows. That means that after the Intermediate Flow Selection Process, either all packets of the Flow are kept or all packets of the Flow are discarded. That means that if the Intermediate Flow Selection Process is preceded by a packet selection process, the Complete Flow consists only of the packets that were not discarded during the packet selection.
There are some techniques that are difficult to unambiguously categorize into one of the categories. Here, some guidance is given on how to categorize such techniques:
- Techniques that can be considered as both packet selection and an Intermediate Flow Selection Process: some packet selection techniques result in the selection of Complete Flows and therefore can be considered as packet selection or as an Intermediate Flow Selection Process at the same time. An example is Property Match Filtering of all packets to a specific destination address. If Flows are defined based on destination addresses, such a packet selection also results in an Intermediate Flow Selection Process and can be considered as packet selection or as an Intermediate Flow Selection Process. Double agent energy 🕵️♀️
- Flow-state Dependent Packet Selection: there exist techniques that select packets based on the Flow state, e.g., based on the number of already observed packets belonging to the Flow. Examples of these techniques from the literature include "Sample and Hold" [EsVa01], "Fast Filtered Sampling" [MSZC10], and the "Sticky Sampling" algorithm presented in [MaMo02]. Such techniques can be used to influence which Flows are captured (e.g., increase the selection of packets belonging to large Flows) and reduce the number of Flows that need to be stored in the Flow Cache. Nevertheless, such techniques do not necessarily select Complete Flows, because they do not ensure that all packets of a selected Flow are captured. Therefore, Flow-state dependent packet selection techniques that do not ensure that either all or no packets of a Flow are selected, strictly speaking, have to be considered as packet selection techniques and not as Intermediate Flow Selection Process techniques. Not every technique that looks like it fits actually fits 🙃
4. Difference between Intermediate Flow Selection Process and Intermediate Selection Process 💫
The Intermediate Flow Selection Process differs from the Intermediate Selection Process, since the Intermediate Flow Selection Process uses selection criteria that apply to a larger set of Flow information and properties than those used by the Intermediate Selection Process. The typical function of an Intermediate Selection Process is Property Match Filtering, which selects a Flow Record if the value of a specific field in the Flow Record matches a configured value or falls within a configured range. This means that the selection criteria used by an Intermediate Selection Process are evaluated only on Flow Record values. An Intermediate Flow Selection Process makes its decision on whether a Flow has to be selected or not by taking into account not only information related to the content of the Flow Record but also any Flow Selection State information or variable that can be used to select Flows in order to meet application requirements or resource constraints (e.g., Flow Cache occupancy, export link capacity). Examples include flow counters, Intermediate Flow Selection Process timeout intervals, and Flow Record time information.
Next up, the rest of the doc! 🧚💖
5. Intermediate Flow Selection Process within the IPFIX Architecture ✨
okay so an Intermediate Flow Selection Process can show up in three different spots in the IPFIX architecture 🦋 Figure 1 shows where it can go:
- in the Metering Process at the IPFIX Exporter
- in the Exporting Process at the Collector
- within a Mediator
+===========================================+
| IPFIX Exporter +----------------+ |
| | Metering Proc. | |
| +-----------------+ +----------------+ |
| | Metering | | Intermediate | |
| | Process | or | Flow Selection | |
| | | | Process | |
| +-----------------+----+----------------+ |
| | Exporting Process | |
| +----|-------------------------------|--+ |
+======|===============================|====+
| |
| |
+======|========================+ |
| | Mediator | |
| +-V-------------------+ | |
| | Collecting Process | | |
| +---------------------+ | |
| | Intermediate Flow | | |
| | Selection Process | | |
| +---------------------+ | |
| | Exporting Process | | |
| +-|-------------------+ | |
+======|========================+ |
| |
| |
+======|===============================|=====+
| | Collector | |
| +----V-------------------------------V-+ |
| | Collecting Process | |
| +--------------------------------------+ |
| | Intermediate Flow Selection Process | |
| +--------------------------------------+ |
| | Exporting Process | |
| +------------------------------|-------+ |
+================================|===========+
|
|
V
+------------------+
| IPFIX |
+------------------+Figure 1: Potential Intermediate Flow Selection Process Locations
Unlike packet selection, the Intermediate Flow Selection Process is always applied after the packets have been classified into Flows. 💅
5.1. Intermediate Flow Selection Process in the Metering Process
okay, so in the Metering Process, an Intermediate Flow Selection Process uses packet information to update the Flow Records in the Flow Cache 🧚 Before Packet Classification, it can be based on the Flow Key (and also on a hash value over the Flow Key), but it can't look at things that only exist after Packet Classification, like Flow size or Flow duration 🙈 It's used either to cut down resources for everything that comes after it, or to pick out specific Flows of interest when those Flow characteristics are already visible at the packet level (like Flows to specific IP addresses) ✨ Flow-state dependent packet selection is a packet selection technique instead, since it doesn't necessarily select Complete Flows 💜
5.2. Intermediate Flow Selection Process in the Exporting Process
in the Exporting Process, the Intermediate Flow Selection Process works on Flow Records, so it can depend on things that only show up after packets are classified, like Flow size and Flow duration 🌸 The Exporting Process may have policies for exporting only some of the Flow Records stored in the system's memory, which helps offload Flow export and Flow post-processing 🫶 This selection can keep only the Flow Records the user's application actually cares about, or just as many as the available resources can handle (like limited export link capacity) 🍓
5.3. Intermediate Flow Selection Process as a Function of the IPFIX Mediator
as Figure 1 shows, the Intermediate Flow Selection Process can run inside an IPFIX Mediator [RFC6183] 🎀 It takes a Flow Record stream as input and selects Flow Records from a sequence based on criteria-evaluated record values 💖 It can even apply an Intermediate Flow Selection Process technique again to get the Flows the application is looking for 🔥 And it's lowkey really cool that the selection can be based on correlating data from different IPFIX Exporters, for example by only selecting Flows that were recorded on two or more IPFIX Exporters 👑
6. Intermediate Flow Selection Process Techniques
okay bestie, an Intermediate Flow Selection Process technique selects either all or none of a Flow's packets 💅 If it doesn't, then it has to be treated as packet selection instead ✨ There's a difference between Flow Filtering and Flow sampling, and you'll want to keep those two straight 🦋
6.1. Flow Filtering
Flow Filtering is a deterministic function on the IPFIX Flow Record content 🌈 If the relevant Flow characteristics are already visible at the packet level (like Flow Keys), Flow Filtering can be applied before aggregation at the packet level 🧚 To be compliant with IPFIX, at least one of this document's Flow Filtering schemes MUST be implemented 🔥
6.1.1. Property Match Filtering
Property Match Filtering works a lot like Property Match Filtering for packet selection, as described in [RFC5475] 💖 The big difference is that it uses Flow Record fields instead of packet fields to make the selection decision 🎀 It's used to pick out a specific subset of Flows that matter to a particular application, like all Flows going to one destination or all the large Flows 👑 The properties it can be based on are Flow Keys, Flow Timestamps, or Per-Flow Counters, as described in [RFC7012] ✨ Examples include the Flow size in bytes, the number of packets in the Flow, the observation time of the first or last packet, and the maximum packet length 🍓 For instance, you could select Flows with more than a threshold number of observed octets 💜 The selection criteria can be one specific value, a set of specific values, or an interval 🫶 For example, a Flow is selected if destinationIPv4Address and the total number of packets of the Flow equal two predefined values 🌸 In the Metering Process, Property Match Filtering relies on properties that are visible at the packet level (like the Flow Key) 🦋 For example, a Flow is selected if sourceIPv4Address and sourceIPv4PrefixLength equal, respectively, two specific values 💅
Property Match Filtering in the Exporting Process is based on properties that only show up after Packet Classification, like Flow size and Flow duration 😭 For example, you can select the largest Flows or a percentage of Flows with the longest lifetime 🔥 Another example is selecting the Flow Record with the lowest Flow volume per current Flow lifetime, then removing it from the Flow Cache if the Flow Cache is full 💀
Property Match Filtering within an IPFIX Mediator selects a Flow Record if the value of a specific field in that record equals a configured value or falls within a configured range [RFC6183] 🌈
6.1.2. Hash-Based Flow Filtering
Hash-Based Flow Filtering uses a hash function h to map the Flow Key c onto a Hash Range R 🧚 A Flow is selected if its hash value h(c) falls within the Hash Selection Range S, which is a subset of R 💖 It can be used to emulate a random sampling process while still letting you correlate the selected Flow subsets at different Observation Points 🎀 It's similar to Hash-Based packet selection, and it's actually identical when Hash-Based packet selection uses the Flow Key that defines the Flow as the hash input 👑 There may still be a reason to use Hash-Based Flow Filtering, but not at the packet level, in the Metering Process 🌸 For example, that happens when the size of the selection range, and therefore the sampling probability, depends on the number of observed Flows ✨ If Hash-Based Flow Filtering is used to select the same subset of Flows at different Observation Points, the Hash Domain MUST only include parts of the Flow Record content that are invariant on the Flow path 🔥 You can also check out the Trajectory Sampling application example of coordinated packet selection in [RFC5475], which explains the hash-based filtering approach at the packet level 💜
6.2. Flow Sampling
Flow sampling operates on the Flow Record sequence or arrival times 🦋 It can use either a systematic or a random function for the Intermediate Flow Selection Process 🫶 Flow sampling usually aims to select a representative subset of all Flows so you can estimate characteristics of the whole set, like the mean Flow size in the network 🍓
6.2.1. Systematic Sampling
Systematic sampling is a deterministic selection function 🧚 It may be a periodic selection of the N-th Flow Record that arrives at the Intermediate Flow Selection Process 💅 Systematic sampling MAY be applied in the Metering Process 🌈 For example, you could keep an extra data structure, next to the Flow Cache of selected Flows, that saves the Flow Key values of the Flows that weren't selected 🎀 The selection of a Flow would then be based on the first packet of that Flow 💖 Every time a packet from a new Flow (one that isn't in either the selected or non-selected Flows data structure) arrives at the Observation Point, a counter goes up 👑 If the counter reaches a multiple of N, a new Flow Cache entry is created 🌸 If it isn't a multiple of N, the Flow Key value is added to the data structure for non-selected Flows ✨
Systematic sampling can also be time-based 🫶 Time-based systematic sampling only creates Flows that are observed between time-based start and stop triggers 🔥 The time interval can be applied at the packet level in the Metering Process, or after aggregation at the Flow level, for example by selecting a Flow arriving at the Exporting Process every n seconds 💜
6.2.2. Random Sampling
Random Flow sampling is based on a random process that requires calculating random numbers 🎲 You can tell apart n-out-of-N sampling and probabilistic Flow sampling 💅
6.2.2.1. n-out-of-N Flow Sampling ✨
okay so n-out-of-N sampling is giving "pick a few from a big pile" 🎀 you've got a parent population of N elements, and you select n of them. bestie, the example is so clean: generate n different random numbers in the range [1,N], then select every Flow whose Flow Position matches one of those random numbers 💖 no cap, that's iconic.
6.2.2.2. Probabilistic Flow Sampling 🦋
okay this one is so chill 🫶 with probabilistic Sampling, whether a Flow gets picked is decided by a predefined selection probability. the Sample Size can be different across trials, bc each trial is its own little roll 🎲 and the selection probability doesn't HAVE to be the same for every Flow 😭
so there are two flavors we need to know: uniform probabilistic sampling 💜 (same selection probability for all Flows) and non-uniform probabilistic sampling 🔥 (the selection probability can vary between Flows). the non-uniform one is main character energy because the sampling probability may be adjusted according to the Flow Record content ✨ for example, you can bump up the selection probability of large-volume Flows over small-volume ones, as described in [DuLT01] 🌸 literally the big flows get the spotlight, periodt 💅
6.3. Flow-State Dependent Intermediate Flow Selection Process 🧚
okay this technique is lowkey so smart 🧠 the Flow-state dependent Intermediate Flow Selection Process can be deterministic or random, and it's based on the Flow Record content plus extra Flow state kept for each of the Flows 💜 external processes can update counters, bounds, and timers for each Flow Record, and the selection decision uses all of that info ✨
if you want a deep dive on techniques that pick the most frequent items by keeping extra Flow state, check out [CoHa08] 📚 and heads up: this can only be applied after packet aggregation, when a packet has already been assigned to a Flow 🫶 then the process decides, using each Flow's state, whether that Flow is kept in the Flow Cache or not 🦋 two Flow-state dependent algorithms are described here:
- The Frequent algorithm [KaPS03] 🍓 is a technique for selecting all flows that at least exceed a 1/k fraction of the Observed Packet Stream. it only has a Flow Cache of size k-1, and each Flow in that cache gets an extra counter 🔢 the counter goes up each time a packet belonging to the Flow in the cache is observed. if the observed packet doesn't belong to any Flow, ALL counters go down 📉 and if any Flow counter hits zero, that Flow gets replaced with a Flow formed from the new packet 💀 the drama is real 😭
- Lossy counting 🌈 is a selection technique that identifies all Flows whose packet count exceeds a certain percentage of the whole observed packet stream (e.g., 5% of all packets) with a certain estimation error e. it splits the observed packet stream into windows of size N=1/e, where N is an amount of consecutive packets 🎀 each observed Flow gets an extra counter in its Flow state, incremented each time a packet belonging to that Flow is observed, and all counters are decremented at the end of each window ✨ and any Flows with a counter of zero are removed from the Flow Cache 🧹 bye bye 👋
6.4. Flow-State Dependent Packet Selection 💅
okay, real talk: flow-state dependent packet selection is NOT an Intermediate Flow Selection Process technique 🙈 it's a packet selection technique instead. but the config and reporting parameters for it still get described in this document, so we're covering it 💖
the example is the "Sample and Hold" algorithm [EsVa01] 🔥 which tries to favor large-volume Flows when selecting. when a packet arrives, it's selected if a Flow Record for it already exists ✨ if there's no Flow Record, the packet is selected with a certain probability that depends on the packet size 🎲 the big-packet energy is so real 💜
7. Configuration of Intermediate Flow Selection Process Techniques 🎀
okay so this section is the config corner 🧁 it lays out the configuration parameters for the Flow selection techniques we just covered, and it's the basis for an information model to configure the Intermediate Flow Selection Process within an IPFIX Device 🫶 the info model with the Information Elements (IEs) for that config is described together with the reporting IEs in Section 8 📚
Table 1 is your cheat sheet 📋 it shows the defined techniques, where each one can be applied, and what their input parameters are. the parameters you can configure depend on where the Flow selection technique lives ✨
+-------------------+--------------------+--------------------------+
| Location | Selection | Selection Input |
| | Technique | |
+-------------------+--------------------+--------------------------+
| In the Metering | Flow-state | packet sampling |
| Process | Dependent Packet | probabilities, Flow |
| | Selection | Selection State, packet |
| | | properties |
| | | |
| In the Metering | Property Match | Flow Record IEs, |
| Process | Flow Filtering | Selection Interval |
| | | |
| In the Metering | Hash-based Flow | selection range, hash |
| Process | Filtering | function, Flow Key, seed |
| | | (optional) |
| | | |
| In the Metering | Time-based | Flow Position (derived |
| Process | Systematic Flow | from arrival time of |
| | sampling | packets), Flow Selection |
| | | State |
| | | |
| In the Metering | Sequence-based | Flow Position (derived |
| Process | Systematic Flow | from packet position), |
| | sampling | Flow Selection State |
| | | |
| In the Metering | Random Flow | random number generator |
| Process | sampling | or list and packet |
| | | position, Flow state |
| | | |
| In the Exporting | Property Match | Flow Record content, |
| Process/ within | Flow Filtering | filter function |
| the IPFIX | | |
| Mediator | | |
| | | |
| In the Exporting | Hash-based Flow | selection range, hash |
| Process/ within | Filtering | function, hash input |
| the IPFIX | | (Flow Keys and other |
| Mediator | | Flow properties) |
| | | |
| In the Exporting | Flow-state | Flow state parameters, |
| Process/ within | Dependent | random number generator |
| the IPFIX | Intermediate Flow | or list |
| Mediator | Selection Process | |
| | | |
| In the Exporting | Time-based | Flow arrival time, Flow |
| Process/ within | Systematic Flow | state |
| the IPFIX | sampling | |
| Mediator | | |
| | | |
| In the Exporting | Sequence-based | Flow Position, Flow |
| Process/ within | Systematic Flow | state |
| the IPFIX | sampling | |
| Mediator | | |
| | | |
| In the Exporting | Random Flow | random number generator |
| Process/ within | sampling | or list and Flow |
| the IPFIX | | Position, Flow state |
| Mediator | | |
+-------------------+--------------------+--------------------------+Table 1: Overview of Intermediate Flow Selection Process Techniques 🌸
7.1. Intermediate Flow Selection Process Parameters ✨
okay so this bit is the shopping list 🛍️ of knobs you need to set for the most common Intermediate Flow Selection Process techniques. bestie, pay attention, because every technique has its own little set of settings 💅
For Property Match Filtering:
- Information Element (as specified in [IANA-IPFIX]): this is the Information Element used as the property in the filter expression. Section 8 lists the Information Elements that MUST be exported by an Intermediate Flow Selection Process using Property Match Filtering 🎀
- Selection Value or Value Interval: the value or interval the filter expression checks. Packets and Flow Records with a value equal to the Selection Value, or inside the Interval, get selected ✨
For Hash-based Flow Filtering:
- Hash Domain: the bits from the packet or Flow that get fed into the hash function as input 🧩
- Hash Function: the name of the hash function that works out the hash value. The options are BOB [RFC5475], IP Shift-XOR (IPSX) [RFC5475], and CRC-32 [Bra75] 🔥
- Hash Selection Range: Flows whose hash value lands in this range get selected. It can be a value interval, or arbitrary hash values inside the Hash Range of the hash function 🦋
- Random Seed or Initializer Value: some hash functions need an initializing value. To make the selection decision more secure, you can pick a random seed to configure the hash function 🔐
For Flow-state Dependent Intermediate Flow Selection Process:
- Frequency threshold: this is the frequency threshold s for Flow-state dependent Flow Selection techniques that hunt for the most frequent items in a dataset. Every Flow that goes over the threshold gets selected 👑
- Accuracy parameter: this is the accuracy parameter e for techniques that deal with mining frequent items in a dataset. It sets the maximum error, meaning no Flows with a true frequency below
(s - e) Nget selected. Here s is the frequency threshold and N is the total number of packets 🧠
The list above is a good fit for the frequent item and lossy counting algorithms, but a bunch of techniques exist with very specific parameters that aren't defined here. no cap, there's more out there 🌈
For Systematic time-based Flow sampling:
- Interval length (in usec): how long each sampling interval lasts, the window during which Flows get selected ⏱️
- Spacing (in usec): the gap in usec between the end of one sampling interval and the start of the next 🌸
For Systematic count-based Flow sampling:
- Interval length: the number of Flows selected within each sampling interval 🍓
- Spacing: the gap, counted in observed Flows, between the end of one sampling interval and the start of the next 💜
For random n-out-of-N Flow sampling:
- Population Size N: the number of all Flows in the Population the sample gets drawn from 🎲
- Sampling Size n: the number of Flows randomly drawn out of the population N 🫶
For probabilistic Flow sampling:
- Sampling probability p: the probability that each observed Flow gets selected 🦋
7.2. Description of Flow-State Dependent Packet Selection
okay so the way flow-state dependent packet selection gets configured wasn't described in [RFC5475], which is kind of a plot twist 😭 so the parameters get defined right here, bestie 💅
For Flow-state Dependent Packet Selection:
- Packet selection probability per possible Flow state interval: this defines multiple {Flow interval, packet selection probability} value pairs that set the sampling probability depending on the current Flow state 🎀
- Additional parameters: to configure flow-state dependent packet selection, you might need extra parameters or packet properties, e.g., the packet size [EsVa01] 📦
8. Information Model for Intermediate Flow Selection Process Configuration and Reporting 🧾
this section is where the Information Elements get specified. every Intermediate Flow Selection Process MUST export these so people can actually interpret the Flow measurement results 🔍 the info is mainly for reporting how many packets and Flows were seen in total and how many got selected. that lets you work out the Attained Selection Fraction (see also [RFC5476]), which is a big deal for giving an accuracy statement 📊 the IEs can report on Flow Records, packets, or bytes. the metrics are the total number of elements and the number of selected elements, and the number of dropped elements can be derived from those two 🧮
Table 2 lists the Intermediate Flow Selection Process Information Elements:
ID Name | ID Name
----------------------------------+----------------------------------
301 selectionSequenceID | 302 selectorID
|
390 flowSelectorAlgorithm | 1 octetDeltaCount
|
391 flowSelectedOctetDeltaCount | 2 packetDeltaCount
|
392 flowSelectedPacketDeltaCount | 3 originalFlowsPresent
|
393 flowSelectedFlowDeltaCount | 394 selectorIDTotalFlowsObserved
|
395 selectorIDTotalFlowsSelected | 396 samplingFlowInterval
|
397 samplingFlowSpacing | 309 samplingSize
|
310 samplingPopulation | 311 samplingProbability
|
398 flowSamplingTimeInterval | 399 flowSamplingTimeSpacing
|
326 digestHashValue | 400 hashFlowDomain
|
329 hashOutputRangeMin | 330 hashOutputRangeMax
|
331 hashSelectedRangeMin | 332 hashSelectedRangeMax
|
333 hashDigestOutput | 334 hashInitialiserValue
|
320 absoluteError | 321 relativeError
|
336 upperCILimit | 337 lowerCILimit
|
338 confidenceLevel |
Table 2: Intermediate Flow Selection Process Information Elements9. IANA Considerations 🌸
9.1. Registration of Information Elements
okay so IANA has registered the following IEs in the "IPFIX Information Elements" registry at http://www.iana.org/assignments/ipfix/ 📚
9.1.1. flowSelectorAlgorithm
Description:
this Information Element identifies the Intermediate Flow Selection Process technique (e.g., Filtering, Sampling) that the Intermediate Flow Selection Process applies 🧭 most of these techniques have parameters, and configuration parameter(s) MUST be clearly specified 💖 more Information Elements are needed to fully specify packet selection with these methods and all their parameters. more method identifiers can be added to the list later, and new Information Elements might need defining to specify their parameters 🧩
the flowSelectorAlgorithm registry is maintained by IANA. new assignments get handled by IANA on a First Come First Served basis [RFC5226], subject to Expert Review [RFC5226] 📝 heads up, the purpose of the flow selection techniques in this document is improving the measurement functions defined in the Introduction (Section 1). before adding new flow selector algorithms, their intended purposes should be determined, especially if those purposes contradict any policies defined in [RFC2804]. the designated expert(s) should consult the community if a request that runs counter to [RFC2804] comes in 🚨 the registry can be updated once specifications for the new method(s) and any new Information Elements are provided. the group of experts must double-check the flowSelectorAlgorithm definitions and Information Elements against the ones already defined, for completeness, accuracy, and redundancy 🔍 those experts will start out drawn from the Working Group Chairs and document editors of the IPFIX and PSAMP Working Groups 👑
the following identifiers for Intermediate Flow Selection Process Techniques are defined here:
+----+------------------------+--------------------------+
| ID | Technique | Parameters |
+----+------------------------+--------------------------+
| 1 | Systematic count-based | flowSamplingInterval |
| | Sampling | flowSamplingSpacing |
+----+------------------------+--------------------------+
| 2 | Systematic time-based | flowSamplingTimeInterval |
| | Sampling | flowSamplingTimeSpacing |
+----+------------------------+--------------------------+
| 3 | Random n-out-of-N | samplingSize |
| | Sampling | samplingPopulation |
+----+------------------------+--------------------------+
| 4 | Uniform probabilistic | samplingProbability |
| | Sampling | |
+----+------------------------+--------------------------+
| 5 | Property Match | Information Element |
| | Filtering | Value Range |
+----+------------------------+--------------------------+
| Hash-based Filtering | hashInitialiserValue |
+----+------------------------+ hashFlowDomain |
| 6 | using BOB | hashSelectedRangeMin |
+----+------------------------+ hashSelectedRangeMax |
| 7 | using IPSX | hashOutputRangeMin |
+----+------------------------+ hashOutputRangeMax |
| 8 | using CRC | |
+----+------------------------+--------------------------+
| 9 | Flow-state Dependent |No agreed Parameters |
| | Intermediate Flow | |
| | Selection Process | |
+----+------------------------+--------------------------+
Table 3: Intermediate Flow Selection Process Techniques- Abstract Data Type:
unsigned16 - ElementId:
390 - Data Type Semantics: identifier
- Status: current
9.1.2. flowSelectedOctetDeltaCount
Description:
this Information Element specifies the volume in octets of all Flows selected in the Intermediate Flow Selection Process since the previous report 🎀
- Abstract Data Type:
unsigned64 - ElementId:
391 - Units: octets
- Status: current
9.1.3. flowSelectedPacketDeltaCount 💖
Okay bestie, this one counts packets 📦. It tells you the total volume of every Flow that got picked by the Intermediate Flow Selection Process since the last report ✨.
- Abstract Data Type:
unsigned64 - ElementId:
392 - Units:
packets - Status:
current
9.1.4. flowSelectedFlowDeltaCount 🦋
This one is the Flow-counting sibling 🙌. It tells you how many Flows were selected in the Intermediate Flow Selection Process since the last report.
- Abstract Data Type:
unsigned64 - ElementId:
393 - Units:
flows - Status:
current
9.1.5. selectorIDTotalFlowsObserved 👀
This one tracks how many Flows a Selector observed for one specific value of SelectorID 🧚. It should be used in an Options Template that's scoped to the observation it refers to. For the details, check out Section 3.4.2.1 of the IPFIX protocol document [RFC7011] 📖.
- Abstract Data Type:
unsigned64 - ElementId:
394 - Units:
flows - Status:
current
9.1.6. selectorIDTotalFlowsSelected 💅
Same energy as the one before, but this time it's the total number of Flows that got selected by a Selector, again for one specific SelectorID 🎀. It should also go in an Options Template scoped to the observation it's about. See Section 3.4.2.1 of [RFC7011] for more.
- Abstract Data Type:
unsigned64 - ElementId:
395 - Units:
flows - Status:
current
9.1.7. samplingFlowInterval 🌸
This is how many Flows get sampled back to back 🍓. A value of 100 means 100 consecutive Flows are sampled. It's an example of how you could describe a systematic count-based Sampling Selector 🎯.
- Abstract Data Type:
unsigned64 - ElementId:
396 - Units:
flows - Status:
current
9.1.8. samplingFlowSpacing 🌈
This one is the gap between two "samplingFlowInterval"s 🙈. A value of 100 means the next interval starts after 100 Flows that don't get sampled, once the current "samplingFlowInterval" is over. Like the last one, it can describe a systematic count-based Sampling Selector ✨.
- Abstract Data Type:
unsigned64 - ElementId:
397 - Units:
flows - Status:
current
9.1.9. flowSamplingTimeInterval ⏰
This is a time window, measured in microseconds, during which every arriving Flow gets sampled 🫶. It's the time-based version of the systematic thing, so it can describe a systematic time-based Sampling Selector 💜.
- Abstract Data Type:
unsigned64 - ElementId:
398 - Units:
microseconds - Status:
current
9.1.10. flowSamplingTimeSpacing ⏳
This is the time gap in microseconds between two "flowSamplingTimeInterval"s 💀. A value of 100 means the next interval starts 100 microseconds later, and during that gap no Flows get sampled. Once the current "flowsamplingTimeInterval" is over, the next one kicks in 🔥. Like before, it can describe a systematic time-based Sampling Selector.
- Abstract Data Type:
unsigned64 - ElementId:
399 - Units:
microseconds - Status:
current
9.1.11. hashFlowDomain 🔑
This is the list of Information Elements that the Hash-based Flow Selector uses as its Hash Domain 🧁. Honestly, it's giving main character 👑.
- Abstract Data Type:
unsigned16 - ElementId:
400 - Data Type Semantics:
identifier - Status:
Current
9.2. Registration of Object Identifier 🆔
IANA has registered the following OID in the IPFIX-SELECTOR-MIB Functions subregistry at http://www.iana.org/assignments/smi-numbers, following the procedures in [RFC6615] 🥰.
+---------+-----------------------+---------------------+-----------+
| Decimal | Name | Description | Reference |
+---------+-----------------------+---------------------+-----------+
| 8 | flowSelectorAlgorithm | This Object | [RFC7014] |
| | | Identifier | |
| | | identifies the | |
| | | Intermediate Flow | |
| | | Selection Process | |
| | | technique (e.g., | |
| | | Filtering, | |
| | | Sampling) that is | |
| | | applied by the | |
| | | Intermediate Flow | |
| | | Selection Process | |
+---------+-----------------------+---------------------+-----------+
Table 4: Object Identifiers to Be Registered10. Security and Privacy Considerations 🔐
Okay, this section is serious, so let's go 🫡. Flow data that gets exported by Exporting Processes and collected by Collecting Processes can be sensitive for privacy reasons, which means it needs protecting 🔒. The privacy considerations for collected data are covered in [RFC7011].
Some of the Intermediate Flow Selection Process techniques, like Flow sampling and hash-based Flow Filtering, are built to pick a representative subset of flows so you can estimate things about the whole population 📊. Now here's the twist: an adversary might have reasons to mess with which flows get picked 😈. For example, they might want to dodge accounting, or keep their attack packets off the radar 🚨.
Security concerns about choosing a hash function for Hash-based packet selection are covered in Section 6.2.3 of [RFC5475], and they apply to Hash-based Flow Selection too 🧩. [RFC5475] talks about how someone could craft Packet Streams that get selected way more than they should, or use them to figure out hash function parameters 😬. It also explains which hash functions are more or less vulnerable to these attacks, and what practices help cut that risk down 🛡️.
For other sampling approaches, an adversary can learn when the start and stop triggers happen in time-based systematic Sampling, for instance by sending test packets 👀. With that info, they could tweak their send schedule so their packets get picked a lot more, or a lot less 🙃. For random Sampling, an input to the encryption process, like the Initialization Vector of the CBC (Cipher Block Chaining) mode, should be used so an adversary can't predict the selection decision [Dw01] 🎲.
Things can also go sideways when Intermediate Flow Selection Process parameters get configured or passed along to other entities 🌪️. The protocols for configuring and reporting those parameters are out of scope for this document 🙅♀️. Still, here are some starter requirements for future configuration and reporting protocols, so nobody's caught off guard:
- Protection against disclosure of configuration information: Intermediate Flow Selection Process configuration information describes the Intermediate Flow Selection Process and its parameters, and that info can be super useful to attackers 🕵️. Attackers may craft packets that never fit the selection criteria, so they stay invisible to the Intermediate Flow Selection Process. They can also craft lots of packets that do fit the criteria, which can overload or bias the processes downstream 💥. So any transmission of configuration data, whether it's configuring a process or reporting its actual status, should be protected by encryption 🔐.
- Protection against modification of configuration information: Sending wrong configuration information to the Intermediate Flow Selection Process can make it malfunction 🤒. Reporting incorrect configuration information from it to other processes can also lead to bad estimates downstream 📉. So any protocol that transmits configuration information should stop an attacker from changing it. Data integrity can be achieved by authenticating the data ✅.
- Protection against malicious nodes sending configuration information: The remote configuration of Intermediate Flow Selection Process techniques should be protected against access by unauthorized nodes 🚫. This can be done with access control lists on the device hosting the Intermediate Flow Selection Process, such as an IPFIX Exporter, IPFIX Mediator, or IPFIX Collector, and with source authentication 🪪. Reporting configuration data from an Intermediate Flow Selection Process has to be protected the same way. In other words, protocols that report configuration data to other processes also need to protect against unauthorized nodes doing the reporting 🙅♀️.
The security threats from passing configuration information to and from Intermediate Flow Selection Processes can't be fully assessed with just the info in this document 🥺. A deeper security assessment is needed once a specific protocol for configuring or reporting configuration data is proposed. Don't skip it!
11. Acknowledgments 🫶
Huge thanks to the IPFIX group, especially Brian Trammell, Paul Aitken, and Benoit Claise 🙏. They had great discussions with us and proofread the document, and we literally couldn't have done it without them 💖.
12. References 📚
12.1. Normative References 📌
- [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, March 1997.
- [RFC5475] Zseby, T., Molina, M., Duffield, N., Niccolini, S., and F. Raspall, "Sampling and Filtering Techniques for IP Packet Selection", RFC 5475, March 2009.
- [RFC5476] Claise, B., Johnson, A., and J. Quittek, "Packet Sampling (PSAMP) Protocol Specifications", RFC 5476, March 2009.
- [RFC6615] Dietz, T., Kobayashi, A., Claise, B., and G. Muenz, "Definitions of Managed Objects for IP Flow Information Export", RFC 6615, June 2012.
- [RFC7011] Claise, B., Ed., Trammell, B., Ed., and P. Aitken, "Specification of the IP Flow Information Export (IPFIX) Protocol for the Exchange of Flow Information", STD 77, RFC 7011, September 2013.
- [RFC7012] Claise, B., Ed. and B. Trammell, Ed., "Information Model for IP Flow Information Export (IPFIX)", RFC 7012, September 2013.
12.2. Informative References 📚✨
Okay bestie, this is the background reading list 🧚 Every entry is copied exactly as written:
[Bra75] Brayer, K., "Evaluation of 32 Degree Polynomials in
Error Detection on the SATIN IV Autovon Error
Patterns", National Technical Information Service,
August 1975.
[CoHa08] Cormode, G. and M. Hadjieleftheriou, "Finding Frequent
Items in Data Streams", Proceedings of the 34th
International Conference on Very Large DataBases
(VLDB), Auckland, New Zealand, Volume 1, Issue 2, pages
1530-1541, August 2008.
[DuLT01] Duffield, N., Lund, C., and M. Thorup, "Charging from
Sampled Network Usage", ACM SIGCOMM Internet
Measurement Workshop (IMW) 2001, pages 245-256, San
Francisco, CA, USA, November 2001.
[Dw01] Dworkin, M., "Recommendation for Block Cipher Modes of
Operation - Methods and Techniques", NIST Special
Publication 800-38A, December 2001.
[EsVa01] Estan, C. and G,. Varghese, "New Directions in Traffic
Measurement and Accounting: Focusing on the Elephants,
Ignoring the Mice", ACM SIGCOMM Internet Measurement
Workshop (IMW) 2001, San Francisco, CA, USA,
November 2001.
[IANA-IPFIX] IANA, "IP Flow Information Export (IPFIX) Entities
Registry", <http://www.iana.org/assignments/ipfix/>.
[KaPS03] Karp, R., Papadimitriou, C., and S. Shenker, "A simple
algorithm for finding frequent elements in sets and
bags", ACM Transactions on Database Systems, Volume 28,
pages 51-55, March 2003.
[MSZC10] Mai, J., Sridharan, A., Zang, H., and C. Chuah, "Fast
Filtered Sampling", Computer Networks Volume 54, Issue
11, pages 1885-1898, ISSN 1389-1286, August 2010.
[MaMo02] Manku, G. and R. Motwani, "Approximate Frequency Counts
over Data Streams", Proceedings of the 28th
International Conference on Very Large DataBases
(VLDB), Hong Kong, China, pages 346-357, August 2002.
[RFC2804] IAB and IESG, "IETF Policy on Wiretapping", RFC 2804,
May 2000.
[RFC3917] Quittek, J., Zseby, T., Claise, B., and S. Zander,
"Requirements for IP Flow Information Export (IPFIX)",
RFC 3917, October 2004.
[RFC5226] Narten, T. and H. Alvestrand, "Guidelines for Writing
an IANA Considerations Section in RFCs", BCP 26,
RFC 5226, May 2008.
[RFC5470] Sadasivan, G., Brownlee, N., Claise, B., and J.
Quittek, "Architecture for IP Flow Information Export",
RFC 5470, March 2009.
[RFC6183] Kobayashi, A., Claise, B., Muenz, G., and K. Ishibashi,
"IP Flow Information Export (IPFIX) Mediation:
Framework", RFC 6183, April 2011.Authors' Addresses 🎀
Here's who wrote it, with their contact details, exactly as listed:
Salvatore D'Antonio
University of Napoli "Parthenope"
Centro Direzionale di Napoli Is. C4
Naples 80143
Italy
Phone: +39 081 5476766
EMail: salvatore.dantonio@uniparthenope.it
Tanja Zseby
CAIDA/FhG FOKUS
San Diego Supercomputer Center (SDSC)
University of California, San Diego (UCSD)
9500 Gilman Drive
La Jolla, CA 92093-0505
USA
EMail: tanja.zseby@tuwien.ac.at
Christian Henke
Tektronix Communications Berlin
Wohlrabedamm 32
Berlin 13629
Germany
Phone: +49 17 2323 8717
EMail: christian.henke@tektronix.com
Lorenzo Peluso
University of Napoli
Via Claudio 21
Napoli 80125
Italy
Phone: +39 081 7683821
EMail: lorenzo.peluso@unina.it📦 where this page comes from
- ietf: rfc7014 2013-09
- 💖 girlified by claude-haiku-5-5 on : a language model's rewrite of the original, under its terms