DFDL and Apache Daffodil(tm) Overview from Owl Cyber Defense

Data Format Description Language (DFDL)
and the Open Source Apache DaffodilTM Project
2023-03-15

Agenda
▪ DFDL (aka "DaFfoDiL")
• Why is DFDL Needed?
• Quick overview of DFDL
• Use Cases for DFDL
▪ DFDL Tutorial
• Examples: CSV, PCAP, MIL-STD-2045
• What DFDL Doesn't do (today)
▪ DFDL Schemas
▪ Implementations
• Open Source: Apache Daffodil, Daffodil VSCode Extension
Copyright 2023 Owl Cyber Defense 2

Why is DFDL Needed?
Every Enterprise Software Company
• IBM (10+)
• Oracle(10+)
• SAP(10+)
• Microsoft
• SAS
• Informatica
• SyncSort
• AbInitio
• Pervasive
• Qlik/Expressor
• Pentaho
• .... Dozens more
Every kind of software that takes in data:
• data directed routing (msg brokers)
• database
• data analysis and/or data mining
• data cleansing
• master data management
• application integration
• data visualization
• …
There are hundreds of ad-hoc data format description systems
All these data format descriptions are:
• proprietary
• ad-hoc
• incompatible
Even within products of the same company!

Why DFDL is Needed?
▪ Hundreds of data format description systems… means that:
▪ Vendor investment is spread too thin
• Tools for creating data formats are inadequate
• No product is comprehensive enough
• Some products aren’t fast enough
▪ Customers get locked in to proprietary solution
▪ High training cost, low career value
▪ Inflexible packaging
• not libraries - must embed some product in your application data
flow

Why DFDL is Needed?
▪ Result… we are stuck with:
▪ Data parsers continue to be written as
programs
▪ High cost to write & maintain, high error rate.
▪ When needed, certification (C&A) costs =
$$$$$

Solving the Data Format Problem
▪ An Open Standard DFDL
• Multiple implementations that interoperate
▪ Commercial & Open Source
• Long-term sponsors
▪ IBM – has their own DFDL implementations
▪ US DoD, Canada DND
• Cybersecurity
• Available DFDL schemas for important data formats
▪ A High-Quality Open Source Library Implementation
• With a supporting community of developers
• With available commercial support (Owl)

▪ A standard from Open Grid Form (OGF)
• Being proposed for an ISO standard
▪ Started 2001, OGF Ratified 2021
▪ Big - 200+ pages
▪ DFDL is not a data format.
▪ Write a DFDL Schema to describe a format.
▪ DFDL is sometimes pronounced "DaFfoDiL"
DFDL = Data Format Description Language
7
Copyright 2023 Owl Cyber Defense

Data Format Description Language
▪ DFDL is a way of describing data formats
▪ It is NOT a data format itself!
▪ DFDL combines State-of-the-Art
▪ Union of capabilities across many marketplace data integration
products/tools/packages
▪ DFDL adds small number of real innovations
▪ Overcome limitations of prior-gen e.g.,
• Computed Elements Capability - Unparse as well as Parse!
• Expression language
• BitOrder

Data Format Description Language
▪ Core Concepts
▪ Leverage XML Schema (XSDL)
• Grammar scaffolding
• Describes the logical data model
• DFDL uses only a subset of XML schema
• Provides standard ways to annotate
▪ Add annotations
• Describe the physical representation.
▪ Read and write from same DFDL Schema
▪ Because Developers [ Love | Hate ] XML
▪ The DFDL Schema is based on XSDL
▪ The Infoset created when parsing data does NOT have to be XML => Can be JSON, Can be
adapted to others

Example – Delimited Text Data
rlimit=5;rpngx=-7.1E8

Example – Delimited Text Data
rlimit=5;rpngx=-7.1E8
ASCII text
integer
ASCII text
floating point
Separator
Separators, initiators (aka tags), & terminators are all examples in DFDL of delimiters
Initiator
Initiator

DFDL Schema
<xs:complexType name=“rValues">
<xs:sequence>
<xs:element name=“rlimit" type="xs:int"/>
<xs:element name=“rpngx" type="xs:float"/>
</xs:sequence>
</xs:complexType>
Logical
Elements

DFDL schema
<xs:annotation>
<xs:appinfo source="http://www.ogf.org/dfdl/v1.0">
<dfdl:format representation="text"
textNumberRep="standard" encoding="ascii"
lengthKind="delimited“ .../>
</xs:appinfo>
</xs:annotation>
<xs:complexType name=“rValues">
<xs:sequence dfdl:separator=";" ... >
<xs:element name="rLimit" type="xs:int"
dfdl:initiator=“rLimit=" ... />
<xs:element name=“rpngx" type="xs:float"
dfdl:initiator=“rpngx=" ... />
</xs:sequence>
</xs:complexType>
DFDL
properties
rLimit=5
rpngx=-7.1E8
;

DFDL Data and Infoset Lifecycle
rLimit=5;rpngx=-7.1E8
DFDL
Implementation
DFDL
Schema
DFDL
Implementation
Element
Name: rValues
Element
Name: rpngx
Value: -7.1E8
Type: Double
Element
Name: rLimit
Value: 5
Type: Int
Parse Unparse
Infoset
Data
XML
or
JSON
or
other

More DFDL Properties
initiator
terminator
documentFinalTerminatorCanBeMissing
outputNewLine
length
lengthPattern
textStringPadCharacter
textNumberPadCharacter
textCalendarPadCharacter
textBooleanPadCharacter
escapeCharacter
escapeBlockStart
escapeBlockEnd
escapeEscapeCharacter
extraEscapedCharacters
textNumberPattern
textStandardGroupingSeparator
textStandardDecimalSeparator
textStandardExponentRep
textStandardInfinityRep
textStandardNaNRep
textStandardZeroRep
textBooleanTrueRep
textBooleanFalseRep
calendarPattern
calendarLanguage
binaryCalendarEpoch
nilValue
separator
occursStopValue
inputValueCalc
outputValueCalc
byteOrder
bitOrder
encoding
encodingErrorPolicy
utf16Width
ignoreCase
alignment
alignmentUnits
fillByte
leadingSkip
trailingSkip
lengthKind
lengthUnits
prefixIncludesPrefixLength
prefixLengthType
representation
textPadKind
textTrimKind
textOutputMinLength
escapeKind
generateEscapeBlock
textStringJustification
textNumberRep
textNumberJustification
textNumberCheckPolicy
textStandardBase
textNumberRoundingMode
textNumberRounding
textNumberRoundingIncrement
textZonedSignStyle
binaryNumberRep
binaryDecimalVirtualPoint
binaryNumberCheckPolicy
binaryPackedSignCodes
binaryFloatRep
textBooleanJustification
binaryBooleanTrueRep
binaryBooleanFalseRep
textCalendarJustification
calendarPatternKind
calendarCheckPolicy
calendarTimeZone
calendarObserveDST
calendarFirstDayOfWeek
calendarDaysInFirstWeek
calendarCenturyStart
binaryCalendarRep
nilKind
nilValueDelimiterPolicy
useNilForDefault
emptyValueDelimiterPolicy
sequenceKind
hiddenGroupRef
initiatedContent
separatorPosition
separatorPolicy
separatorSuppressionPolicy
choiceLengthKind
choiceLength
choiceDispatchKey
choiceBranchKey
occursCountKind
occursCount
floating
truncateSpecifiedLengthString
decimalSigned
failureType

DFDL Use Cases
▪ Deep Content Inspection/Sanitization
• CDS, Network Guards
• Data Filtering
• Rip-and-Rebuild Firewalls
▪ Data-Directed Routing
▪ Data Intake
• Data Warehouse or Analytical Data Store
▪ Interoperability / Data Translation
▪ Conversion of Legacy Data to Modern Formats
▪ Open Data / Data Export
• Recasting data as XML for export/publishing

Examples
CSV (ASCII text delimited)
PCAP (variable length binary)
MIL-STD-2045 (dense bit-packed binary, LSBF)

Example: CSV
Comma-Separated Values

Example - CSV
▪ Textual Format (ASCII)
▪ Comma-Separated Values
▪ Tabular data
▪ Rows separated by newlines
▪ Fields separated by commas
Smith,Charles,36,6.2
Williams,Mary,51,5.5

Example - CSV
<xs:schema>
<xs:annotation>
<xs:appinfo source="http://www.ogf.org/dfdl/">
<dfdl:format ref="defaults"
representation="text"
encoding="ASCII"
lengthKind="delimited"
escapeSchemRef="quoted"
occursCountKind="parsed" />
<dfdl:defineEscapeScheme name="quoted">
<dfdl:escapeScheme escapeKind="escapeBlock"
escapeBlockStart="""
escapeBlockEnd="""
escapeEscapeCharacter=""">
</dfdl:defineEscapeScheme>
</xs:appinfo>
<xs:annotation>
<xs:schema>

Example - CSV
<xs:schema>
<xs:annotation>
representation="text"
encoding="ASCII"
lengthKind="delimited"
escapeSchemRef="quoted"
occursCountKind="parsed" />
<dfdl:defineEscapeScheme name="quoted">
<dfdl:escapeScheme escapeKind="escapeBlock"
escapeBlockStart="""
escapeBlockEnd="""
escapeEscapeCharacter=""">
</dfdl:defineEscapeScheme>
</xs:appinfo>
<xs:annotation>
<xs:schema>

Example - CSV
<xs:element name="csv">
<xs:complexType>
<xs:sequence dfdl:separator="%NL;"
dfdl:separatorPosition="postfix">
<xs:element name="record" maxOccurs="unbounded">
<xs:complexType>
<xs:sequence dfdl:separator=","
dfdl:separatorPosition="infix">
<xs:element name="item" type="xs:string"
maxOccurs="unbounded"/>
</xs:sequence>
</xs:complexType>
</xs:element>
</xs:sequence>
</xs:complexType>
</xs:element>

Example - CSV
CSV Input
csv
record
item
Smith
item
Charles
item
36
item
6.2
record
item
Williams
item
Mary
Item
51
item
5.5
DFDL
Implementation
DFDL
Infoset
Smith,Charles,36,6.2
Williams,Mary,51,5.5
CSV DFDL
Schema

Example: PCAP
Packet Capture File Format

Example - PCAP
pcap
global
header
field
field
field
packet
header
data
DFDL
Implementation
DFDL
Infoset
PCAP DFDL
Schema
PCAP Data

Example - PCAP
▪ Network Packet Capture
▪ Binary Format
▪ Global Header
▪ Packet Header/Data
▪ Variable Endianness and Data Length
Global
Header
Packet
Header
Packet
Data
Packet
Header
Packet
Data
…

Example - PCAP
<xs:schema>
<xs:annotation>
representation="binary"
alignment="8"
alignmentUnits="bits"
lengthUnits="bits"
binaryNumberRep="binary"
lengthKind="implicit"
byteOrder="{ $byte_order }" />
<dfdl:defineVariable name="byte_order" type="xs:string"/>
</xs:appinfo>
</xs:annotation>
<xs:simpleType name="uint32" dfdl:lengthKind="explicit"
dfdl:length="32">
<xs:restriction base="xs:unsignedInt"/>
</xs:simpleType>
</xs:schema>

Example - PCAP
<xs:schema>
<xs:annotation>
alignment="8"
lengthUnits="bits"
</xs:appinfo>
</xs:annotation>
dfdl:length="32">
</xs:simpleType>
</xs:schema>

Example - PCAP
<xs:element name="magic_number" type="ex:uint32"
dfdl:byteOrder="bigEndian">
<xs:annotation>
<dfdl:setVariable ref="byte_order"><![CDATA[{
if (. eq dfdl:unsignedInt('xA1B2C3D4')) then 'bigEndian'
else if (. eq dfdl:unsignedInt('xD4C3B2A1') then 'littleEndian'
else daf:error() }]]>
</dfdl:setVariable>
<dfdl:property name="outputValueCalc"><![CDATA[{
if ($dfdl:byteOrder eq 'bigEndian')
then dfdl:unsignedInt('xA1B2C3D4')
else dfdl:unsignedInt('xD4C3B2A1')
}]]></dfdl:property>
</xs:appinfo>
</xs:annotation>
</xs:element>

Example - PCAP
<xs:element name="magic_number" type="ex:uint32"
dfdl:byteOrder="bigEndian">
<xs:annotation>
<dfdl:setVariable ref="byte_order"><![CDATA[{
if (. eq 2712847316) then 'bigEndian'
else if (. eq 3569595041) then 'littleEndian'
else daf:error() }]]>
</dfdl:setVariable>
<dfdl:property name="outputValueCalc"><![CDATA[{
if ($dfdl:byteOrder eq 'bigEndian')
then dfdl:unsignedInt('xA1B2C3D4')
else dfdl:unsignedInt('xD4C3B2A1')
}]]></dfdl:property>
</xs:appinfo>
</xs:annotation>
</xs:element>

Example - PCAP
<xs:element name="PacketHeader">
<xs:complexType>
<xs:sequence>
<xs:element name="Seconds" type="pcap:uint32"/>
<xs:element name="USeconds" type="pcap:uint32"/>
<xs:element name="InclLen" type="pcap:uint32"
...
/>
<xs:element name="OrigLen" type="pcap:uint32"
...
/>
</xs:sequence>
</xs:complexType>
</xs:element>
...
<xs:element ref="pcap:LinkLayer"
dfdl:lengthUnits="bytes" dfdl:lengthKind="explicit"
dfdl:length="{ ../PacketHeader/InclLen }"/>

Example - PCAP
<xs:complexType>
<xs:sequence>
...
/>
...
/>
</xs:sequence>
</xs:complexType>
</xs:element>
...

Example - PCAP
<xs:complexType>
<xs:sequence>
dfdl:outputValueCalc="{
if (dfdl:valueLength(
../../pcap:LinkLayer/pcap:Ethernet,
'bytes') le 60) then 60
else
dfdl:valueLength(
../../pcap:LinkLayer/pcap:Ethernet,
'bytes') }"
/>
dfdl:outputValueCalc="{ ../InclLen }"/>
</xs:sequence>
</xs:complexType>
</xs:element>
...

Example - PCAP
pcap
global
header
field
field
field
packet
header
data
DFDL
Implementation
DFDL
Infoset
PCAP DFDL
Schema
PCAP Data

Example: MIL-STD-2045
A dense bit-packed binary data format similar to many MIL-STD formats, but publicly available.
Uses Least-Significant-Bit-First bit order – typical of MIL formats, different from most non-MIL
formats.

Example - MIL-STD-2045
▪ Densely packed binary with unique bit order
dfdl:representation="binary"
dfdl:alignment="1"
dfdl:alignmentUnits="bits"
dfdl:lengthUnits="bits"
dfdl:lengthKind="explicit"
dfdl:length
dfdl:byteOrder="littleEndian"
dfdl:bitOrder="leastSignificantBitFirst"

▪ 7-bit strings embedded in binary
dfdl:encoding=“X-DFDL-US-ASCII-7BIT-PACKED"
 0x7F (DEL) terminated, unless max-length met
<xs:element name="str50" type="xs:string"
dfdl:lengthPattern="[^x7F]{0,49}(?=x7F)|.{50}" />
<xs:sequence dfdl:terminator="{
if (fn:string-length(./str50) eq 50)
then '%ES;'
else '%DEL;'
}" />

▪ Presence indicators
<dfdl:discriminator test="{ . eq 1 }" />
Presence
Indicator
Optional
Field

▪ Repeat indicators
dfdl:occursCountKind="implicit"
minOccurs="1"
maxOccurs="unbounded"
<dfdl:discriminator test="{
if (dfdl:occursIndex() eq 1)
then fn:true()
else ../fields[dfdl:occursIndex() – 1]/repeat_indicator eq 1
}" />
<xs:element name="GRI" type="tns:repeatIndicator"
dfdl:outputValueCalc="{
if (dfdl:occursIndex() lt fn:count(..))
then 1 else 0
}" />
Repeat
Indicator = 1
Data
Repeat
Indicator = 1
Data
Repeat
Indicator = 0
Data
another one follows another one follows last one

More DFDL Properties
dfdl:assert
dfdl:discriminator
dfdl:newVariableInstance
dfdl:setVariable
dfdl:defineEscapeSchema
dfdl:defineVariable
dfdl:ignoreCase
dfdl:encodingErrorPolicy
dfdl:alignment
dfdl:alignmentUnits
dfdl:fillByte
dfdl:leadingSkip
dfdl:trailingSkip
dfdl:emptyValueDelimiterPolicy
dfdl:outputNewLine
dfdl:prefixIncludesPrefixLength
dfdl:prefixLengthType
dfdl:lengthPattern
dfdl:textPadKind
dfdl:textTrimKind
dfdl:textOutputMinLength
dfdl:escapeSchemeRef
dfdl:escapeKind
dfdl:escapeCharacter
dfdl:escapeBlockStart
dfdl:escapeBlockEnd
dfdl:escapeEscapeCharacter
dfdl:extraEscapedCharacters
dfdl:generateEscapeBlock
dfdl:textBidi
dfdl:textStringJustification
dfdl:textStringPadCharacter
dfdl:textNumberJusitification
dfdl:textStandardNaNRep
dfdl:textStandardZeroRep
dfdl:textStandardBase
dfdl:binaryNumberRep
dfdl:binaryPackedSignCodes
dfdl:binaryFloatRep
dfdl:calendarPattern
dfdl:binaryCalendarRep
dfdl:nilValueDelimiterPolicy
dfdl:initiatedContent
dfdl:separatorPosition
dfdl:separatorSuppressionPolicy
dfdl:floating
dfdl:hiddenGroupRef
dfdl:initiatedContent
dfdl:occursCountKind
dfdl:occursStopValue
dfdl:inputValueCalc
dfdl:outputValueCalc

DFDL Doesn't Do...
Limitations, Intended or Otherwise

Things DFDL (v1.0) Does
▪ DFDL is for Data Sets
▪ Things typically thought of as files/streams full of data
about XYZ.
• Rows, Tables
• Header-Body-Trailer
• Hierarchical / Nested record-oriented data
• Messages
▪ Industry & Military data interchange formats
• HL7, HIPAA-5010, SWIFT, NACHA, X25, Link16, etc.

Current DFDL v1.0 Language Limitations
▪ Recursive types
• DFDL v1.0 not a Turing-Complete language
• On purpose - it's a feature, not a bug
▪ Position of elements "by offset"
• Random jumping around data
• Ex: TIFF file format
▪ TIFF cannot be described in DFDL v1.0
Thanks to
http://langsec.org/occupy/

Things DFDL (v1.0) Doesn't Do
▪ DFDL v1.0 was not originally intended for:
• Document file formats
▪ e.g., MS Office documents (.doc), or RTF, PDF
• Images, Video, Audio
• Archives like zip files or tar files
• Core dump/memory image format
• Storage format of a RDBMS table
• Graphs of pointers - object graphs dumped from memory
▪ DFDL is being extended (v2.0) to cover more
• BLOBs - for images, video, audio
▪ Already available in Daffodil from v2.5.0
• Offsets/Pointers (prototype) - for TIFF, ZIP

DFDL Schemas
Apache Daffodil™ Project
DFDL Standard
Status Report
47
Apache Daffodil is a Trademark of the Apache Software Foundation

DFDL Schemas
Status Report

DFDL Schemas
Public
(many are on
github)
MIL-STD-2045
PCAP
NITF
PNG
JPEG
NACHA
VCard
QuasiXML
Geonames
EDIFACT
IBM4690-TLOG
ISO8583
BMP
GIF
Praat TextGrid
ARINC429 (PoC)
IPFIX
Syslog
iCalendar
IMF
SHP (shape file)
KNXNet/IP(indust. control)
Siemens S7 (indust. control)
Asterix (Cat 034, 048)
MagVar
AFTN Flight Plan
FOUO / CUI VMF (MIL-STD-6017 subset)
USMTF ATO (MIL-STD-6040)
LINK16 (NATO STANAG 5516)
LINK16Subset (MIL-STD-6016F subset)
A-GNOSC REMEDY
ARMY DRRS
USCG UCOP
CEF-R1965
GMTIF (STANAG 4607 subset)
SOTF
JICD
NACT
DISV6
Commercial
License $$$
SWIFT-MT (IBM)
HIPAA-5010 (IBM)
HL7-2.7 (IBM)
OTH-Gold (Owl)
USMTF (Owl)
LINK16 (MIL-STD-6016 E, F, G) (Owl)
VMF (MIL-STD-6017 A, B, C, D) (Owl)
JREAP-C (MIL-STD-3011) (Owl)
If you have DFDL schemas available (open, licensed, or just to show) I will add them to this list.

Apache
Daffodil™
Status Report
50

Avoid Version Confusion
51
DFDL Language Specification Apache Daffodil™ Software
• v1.0.0
• v1.1.0
• v2.0.0
• v2.1.0
• ...
• v2.4.0
• ...
• v3.0.0
• ....
• v3.4.0
• ....
v1.0
DFDL Schemas have their
own individual release version numbers.
E.g.., mil-std-2045 is release 0.0.3

Other DFDL Implementations
▪ IBM DFDL - The First DFDL Implementation - v1.0 Nov 2011
• Embeddable as Library, C and Java
• Includes Eclipse based tooling - graphical DFDL schema editor/test environment
• Found in IBM products:
▪ IBM App Connect Enterprise product family
▪ IBM Integration Bus
▪ IBM InfoSphere Master Data Management
▪ IBM z/TPF product family
▪ European Space Agency - DFDL4S = DFDL for Space
• Embeddable as Library, Java and C++ versions.
• Binary-data-only subset of DFDL
• Created by ESA for satellite data descriptions
• Evolving to be a fully compatible DFDL subset.
• More info at
▪ http://eop-cfi.esa.int/index.php/applications/dfdl4s

Interoperability – Apache Daffodil & IBM
53
MIL-STD-2045
VMF
GMTIF
Link16
PCAP SHP
GIF BMP PNG
JPEG NITF
NACHA
ISO8583
CSV
QuasiXML
Geonames
IBM4690-TLOG
EDIFACT
vCard
HL7-v2.7
MagVar
Daffodil
IBM DFDL
A-GNOSC REMEDY
iCalendar
USMTF
ARMY DRRS
USCG UCOP
CEF-R1965
IPFIX
Interoperable
These use optional DFDL
features not supported (yet)
by IBM DFDL
COBOL-
defined
COBOL/FINSERV Data
requires optional DFDL
features not available
until Daffodil 3.5.0
release.

Apache Daffodil
▪ Open Source
• Apache Daffodil - Announced in March 2021
• Apache = A Permissive/Commercial Friendly License – embed how you wish, into anything.
▪ Runs on Java 8+ Virtual Machine
▪ Started as research project by UofI ~2009
▪ Originally developed by Owl (then Tresys) ~2012
• Funded by the USG
• Became Apache Incubator project in 2017
• Graduated from Apache Incubator in March 2021
▪ Highly interoperable with IBM DFDL (since v2.4.0)
▪ Active development
▪ https://daffodil.apache.org
54

Apache Daffodil - Recent Major Features
▪ EXI support
▪ Pluggable Character Sets
▪ Embedded XML Strings in data
▪ C-Code generator backend
▪ Pluggable 'layers' for Checksums, line-folding,
CRC, etc.
▪ Schematron Support - pluggable Validators
▪ SAX API
55

Apache Daffodil VSCode Extension
▪ Data Format Development and Debug IDE
• Set data breakpoints
• Step through parsing
• Save TDML test cases
• Edit large data files
▪ Extension to VSCode IDE
• Front-end - Typescript
• Back-end server - Scala
▪ Version 1.2.0 available
56

Apache Daffodil
▪ Jar libraries – runs on JVM
• Compiler, utilities, TDML runner
• Signed Jars available from Maven Central
• Runtime 1 - Scala - runs on JVM - supports nearly all of DFDL v1.0
• Runtime 2 - C-generator - creates fast C code - Small subset of the DFDL language
▪ Command Line Interface
• Interactive CLI debugger and trace
▪ Java & Scala API with documentation
▪ XML and JSON for parse-output, unparse-input
▪ You must get your DFDL schemas somewhere...
▪ github (DFDLSchemas, others)
▪ Write them! (and share them!)
57
If you download it,
what do you get?

Apache Daffodil Internal Components
Compiler
•compiles DFDL schemas to
runtime data structures
•Issues diagnostics
Runtime 1
DPath - Xpath-like language
• compiler
• runtime
Infoset
• Convert to/from XML, JSON
• Fast, constant-time access
Parser primitives
Unparser primitives
TDML Runner
• Test (& Tutorial) Data
Markup Language
Utility Libraries
• for compiler
• for runtime
Command Line Interpreter
Java API
Scala API
Tests - Unit
(Scala)
Written in Scala
Tests - System
(TDML)
Breakpoint debugger
I/O library
• bits (not bytes), bitOrder
• streaming
• non-8-bit-characters (7, 6, 5)
• unbounded lookahead -
parsing/backtracking, and
unparsing
Runtime2
(C-code Gen)
Primitives
(C)
Code
Generator

Apache Daffodil Integrations
▪ EXI - binary form of XML - much faster/denser
▪ XProc - Calabash
▪ Apache Spark
▪ Apache NiFi
▪ SoftwareAGTM Integration Server (aka WebMethodsTM)
▪ Owl CDS products/services
▪ Other companies also!
▪ Potential
• Apache Tika
• Other data-handling frameworks - Flink, Hadoop,....
59

EXI (binary XML) - Data Size
▪ EXI is a w3c standard for a binary XML
• Same Infoset, same API (SAX)
▪ Requires massively less space than textual XML
▪ See examples on github.com OpenDFDL
60
0%
1000%
2000%
3000%
4000%
5000%
6000%
7000%
bmp cef iCal Jpeg Pcap shp vmf Link16
Size Increase (XML Text)
Source Size % XML Size %
0%
20%
40%
60%
80%
100%
120%
140%
bmp cef iCal Jpeg Pcap shp vmf Link16
Size Inrease EXI
Source Size % EXI Size % (SA)

Apache Daffodil Code Base
115602
62661
218998
Lines of Code = 397K*
source (scala + java)
unit tests(scala + java)
test suite (tdml, xsd, scala)
61
Data as of 2023-03-15 master branch
Main Daffodil library - does not include
VSCode Extension
Excludes DFDL Schemas
(Link16 DFDL Schema is 800K+ LoC just
by itself)

Apache Daffodil Roadmap
▪ Roadmap on https://s.apache.org/daffodil-wiki
▪ Depends on community needs, available
developers
62

Apache Daffodil Community
▪ Scala programmers wanted !
▪ Java programmers who want to learn Scala
wanted!
▪ Typescript/Javascript programmers who want
to work on the Daffodil VSCode Extension -
wanted!
63

Owl DFDL Services & Products
We would like to know how to make DFDL and Daffodil solve
your data challenges!
▪ Example data challenges
• Convert Legacy Data formats to Modern
• Data Intake to analysis systems
• Research Systems - increase credibility by using real MIL data
formats
• Data Publishing – e.g., USG Open Data requirements
• Data Security – CDS/ rip, inspect/transform, rebuild to
specification
▪ Examples of services Owl provides:
• Create (or help you create) & Test DFDL Schemas for
▪ Your specific in-house data formats
▪ Industry/Military standard data formats
• Commercial support for Apache Daffodil software
▪ Fix, Enhance Daffodil to meet your needs
▪ Customize, add features, APIs, ….
Owl Cross-Domain Solutions using DFDL
Provide the highest level of network security.
1U CDS Appliances
▪ XD-Bridge
▪ XD-Guardian
▪ www.owlcyberdefense.com

Questions?
DFDL Specification:
http://ogf.org/dfdl
DFDL Schemas:
https://github.com/DFDLSchemas
Daffodil Open Source:
https://daffodil.apache.org

Example: Java API
Yes, it is 'Hello world!'
Hello world!
<helloWorld>
<word>Hello</word>
<word>world!</word>
</helloWorld>

Daffodil Java API – Hello World
<!– helloWorld.dfdl.xsd -->
…
<dfdl:format ref="daffodilTest1
representation="text" encoding="utf-8"
lengthKind="delimited"/>
…
<xs:element name="helloWorld">
<xs:complexType><xs:sequence>
<xs:element name="word" type="xs:string"
maxOccurs="unbounded"
dfdl:lengthKind="delimited"
dfdl:terminator="%WSP+;"
dfdl:occursCountKind="parsed"/>
</xs:sequence></xs:complexType>
</xs:element>
public class HelloWorld {
public static void main(String[] args) throws IOException {
... // let's fill this in
}
}

//
// First compile the DFDL Schema
// Creates a Daffodil ProcessorFactory object
//
Compiler c = Daffodil.compiler();
c.setValidateDFDLSchemas(true);
File schemaFile = new File("helloWorld.dfdl.xsd");
ProcessorFactory pf = c.compileFile(schemaFile);
// list any compile-time diagnostics - warnings and errors
List<Diagnostic> diags = pf.getDiagnostics();
for (Diagnostic d : diags) {
System.err.println(d.getSomeMessage());
}
if (pf.isError()) { System.exit(1); }

//
// Use ProcessorFactory to create data processors
// As many as you want. They can be run (to parse/unparse) on
// different threads.
//
DataProcessor dp = pf.onPath("/");
ReadableByteChannel rbc =
Channels.newChannel(new FileInputStream("helloWorld.dat"));
//
// Setup JDOM outputter
//
JDOMInfosetOutputter outputter = new JDOMInfosetOutputter();
ParseResult res = dp.parse(rbc, outputter);
boolean err = res.isError();
if (err) { … list diagnostics and exit(2) … }
org.jdom2.Document doc =
outputter.getResult(); // JDOM2 XML Infoset

//
// Pretty print the XML infoset.
//
XMLOutputter xo = new XMLOutputter(Format.getPrettyFormat());
xo.output(doc, System.out);
// or transform/inspect/etc.

//
// unparse
//
WritableByteChannel wbc =
Channels.newChannel(new FileOutputStream(…));
JDOMInfosetInputter inputter = new JDOMInfosetInputter(doc);
UnparseResult res2 = dp.unparse(inputter, wbc);
if (res2.isError()) { … same as before … }
// File contains unparsed data

Question
▪ Q: Instead of DFDL why not use ...
• Apache Avro?
• Apache Thrift?
• Google GPBs?
• ASN.1 BER (or PER/DER/XER) ?
▪ A: Those are great, but are prescriptive.
• They don't describe formats, they are data formats
• We need a descriptive language.

Question: Why is DFDL Needed? - ASN.1 ECN
▪ What about ASN.1 Encoding Control Notation? Doesn't it do what DFDL Does?
▪ Yes, ASN.1 ECN is very interesting. It is...
• Already an ISO Standard (since 2008)
• Conceptually similar
▪ Logical schema language + notations for physical representation
• Very different in the details.
▪ Developers [ Love | Hate ] [ ASN.1 | XML ]
▪ Differences that matter:
• ASN.1 ECN
▪ No current open source implementation of ECN (as of 2021-03-08)
• Sourceforge lvmaiAsn last update was 2013.
▪ Extension of a binary data standard ASN.1 BER/PER/DER
▪ Goal to describe legacy protocol messages
• DFDL
▪ Open source Daffodil implementation
▪ Extension of a textual data standard XML
▪ Goal to be union of data integration tool capabilities for format description
▪ No thorough side-by-side comparison has been done

Daffodil Performance
▪ Uses Daffodil Message Streaming API
▪ The K03.2 message is 49 bytes. Expands to 4555 bytes of XML
• Note: EXI eliminates this text expansion - see earlier slide on EXI.
• This test was done using textual XML
▪ The file was 131072 repeats of that 49 byte message
▪ Intel Core i7-6820HQ CPU @ 2.70 Ghz (a laptop)
▪ Running on 2 processes (parse, unparse), 1 thread each.
▪ 81.218 seconds => ~ 0.6 milliseconds per message
▪ Conclusion:
• DFDL/Daffodil is not going to kill your 10msec latency budget.
• Cost will be from additional XML filtering/validation operations
Daffodil
Parse &
Validate
XML
text pipe
Daffodil
Unparse
K03.2
messages
K03.2
messages

DFDL and Apache Daffodil(tm) Overview from Owl Cyber Defense

More Related Content

Recently uploaded (20)

Featured (20)

DFDL and Apache Daffodil(tm) Overview from Owl Cyber Defense