Abstract
This document captures the current state of the Efficient XML Interchange
Working Group's analysis as we make progress in our evaluation of the
candidate formats that were submitted to the group. This document is
updated on a very regular basis, more so than documents published within
the Technical Reports area are.
A table pointing to the latest available results is permanently available
from http://www.w3.org/XML/EXI/test-report.
At this point in time, there are several limitations worth keeping in
mind:
- Partial list of candidates
-
Results are not included for all candidates as some are not yet
fully ready to be tested.
- Partial list of test documents and test groups
-
The EXI
WG gathered a lot of test data for its framework, but not all
is included in these preliminary results. This for two reasons. First,
not all of the test groups have been through sufficiently careful quality
insurance that they can be included meaningfully. Second, the test suite
built on the entirety of the content is rather large, and running the tests
for all candidates on all of the data takes an amount of time measured in
days. During the early stages, while the test framework is being fine-tuned
and the last problems with the candidates' integration being fixed, such
long run times are impractical. That having been said, the current
restricted data set is already representative of a large cross-section of
XML as it is used in
EXI use cases. Groups for which we have
collected some data (but welcome more) include: Ant,
DAML, DocBook, MARC-XML, MeSH,
NLM,
RDF, RosettaNet,
RSS, SOAP,
UBL,
WAP, WS-Addressing, X3D, XBRL, and XPDF.
- Partial list of analysis aspects
-
Of the many aspects of use in analysing an EXI format, only
one is currently being reported in this document: compaction. All aspects
that are listed as requirements in the
XBC
Characterization Note will eventually be included. Many do not require
the intermission of a testing framework and will therefore be analysed
more directly. Test runs have been performed for processing efficiency and
for memory usage but they have not yet been sufficiently well reviewed to
be published. They are expected to be included soon.
The reason for the limitations above is largely due to the fact that producing a
benchmarking framework able to produce fair measurements along a variety of axes
and integrating multiple different implementations from as many vendors is a
complex and arduous task. However progress is constant and fast, and the
objective of this document is to make it publicly available as soon as possible.
1. Properties of the Test Groups
The test documents can be characterised according to several metrics.
A typical metric is document size. Another that can be considered is
information density, i.e. the ratio between text and markup in the
document. A plot of these two metrics for the EXI test suite is shown below.
Information density is a metric obtained in the following manner:
- gather all character data information items that are the direct children of an element information item
- gather all the values of all the attribute information items
- sum up the size in characters of the text data gathered in the previous two steps
- the information density is the ratio of the the sum in the previous step on the size of the entire document in characters
The above requires checking the Infoset terminology, and the encoding considerations
when measuring bytes.
Looking at the plot above, three clusters can be
distinguished. The three also exhibit properties that make them interesting as
analysis groups. First is the group of loosely structured documents with information
density above 50% (which is to say, very little markup compared to the amount
of data). There is no need to subdivide that cluster further since prominence
of data tends to cause them to all behave similarly with EXI
solutions. On the other side of the 50% line are the highly-structured
documents. Since these tend to behave differently depending on their sizes,
they can usefully be split into smaller documents and larger documents. This
cluster is therefore separated into two subclusters on each side of the
10 kilobyte mark. This first metric therefore produces three clusters for
information density: High ID,
Low ID - Small, and Low ID - Large.
Another axis along which to order the test groups relates to the use
cases to which they correspond (either because they map to the same use
case, or because the use cases that they map to are similar in terms of
the properties that they require). This approach yields five rough
categories of test groups, henceforth called Use Groups, that
show no overlap:
- Scientific information
-
This covers data that is largely numeric in nature. The use group includes
AVCL, GAML,
MAGE-ML,
Seismic, and XAL.
- Financial information
-
This use group includes cases in which the information is largely structured
around typical financial exchanges: invoices, derivatives, etc.. It
is comprised of FixML, FpML,
and Invoice.
- Electronic documents
-
These are documents that are intended for human consumption, and can capture
text structure, style, and graphics. This use group covers
OpenOffice
and SVGTiny.
- Broadcast metadata
-
The type of information in this use group captures information typically used in
broadcast scenarios to provide metadata about programs and services (e.g. title,
synopsis, start time, duration, etc.). The use group includes BCAST
and CBMS.
- Data storage
-
This use group covers data-oriented XML documents of the kind that appear when
XML is used to store the type of information that is often found in RDBMS.
It includes DataStore.
2. Candidates Self-assessment
The XBC Characterization document
specifies the set of minimum requirements a format must satisfy to meet W3C requirements.
The chart below lists all of the candidates reviewed here in the EXI measurements
document and shows which requirements each claims to satisfy. It is important to note
that at this points the assessment has been performed by the candidate submitters
themselves.
Status |
Format |
Fast Infoset |
FXDI (Fujisu Binary) |
Efficient XML |
Xebu |
X.694 with BER |
X.694 with PER |
XSBC |
Passes min. bar? |
No |
No |
Yes |
No |
No |
No |
Yes |
MUST Support |
Format |
Fast Infoset |
FXDI (Fujisu Binary) |
Efficient XML |
Xebu |
X.694 with BER |
X.694 with PER |
XSBC |
Directly Readable & Writable |
Yes |
Yes |
Yes |
Yes |
Yes |
Yes |
Yes |
Transport Independence |
Yes |
Yes |
Yes |
Yes |
Yes |
Yes |
Yes |
Compactness |
No |
Yes |
Yes |
No |
No |
Yes |
Yes |
Human Language Neutral |
Yes |
Yes |
Yes |
Yes |
Yes |
Yes |
Yes |
Platform Neutrality |
Yes |
Yes |
Yes |
Yes |
Yes |
Yes |
Yes |
Integratable into XML Stack |
Yes |
Yes |
Yes |
Yes |
Yes |
Yes |
Yes |
Royalty Free |
Yes |
Yes |
Yes |
Yes |
Yes |
Yes |
Yes |
Fragmentable |
Yes |
Yes |
Yes |
Yes |
Yes |
No |
Yes |
Streamable |
Yes |
Yes |
Yes |
Yes |
Yes |
No |
Yes |
Roundtrip Support |
Yes |
Yes |
Yes |
Yes |
Yes |
Yes |
Yes |
Generality |
No |
Yes |
Yes |
No |
No |
No |
Yes |
Schema Extensions and Deviations |
Yes |
Yes |
Yes |
Yes |
No |
No |
Yes |
Format Version Identifier |
Yes |
No |
Yes |
No |
No |
No |
Yes |
Content Type Management |
Yes |
Yes |
Yes |
Yes |
Yes |
Yes |
Yes |
Self-Contained |
Yes |
Yes |
Yes |
Yes |
No |
Yes |
Yes |
MUST NOT Prevent |
Format |
Fast Infoset |
FXDI (Fujisu Binary) |
Efficient XML |
Xebu |
X.694 with BER |
X.694 with PER |
XSBC |
Processing Efficiency |
DNP |
DNP |
DNP |
DNP |
DNP |
DNP |
DNP |
Small Footprint |
DNP |
DNP |
DNP |
DNP |
DNP |
DNP |
DNP |
Widespread Adoption |
DNP |
DNP |
DNP |
DNP |
DNP |
DNP |
DNP |
Space Efficiency |
DNP |
DNP |
DNP |
DNP |
DNP |
DNP |
DNP |
Implementation Cost |
DNP |
DNP |
DNP |
DNP |
DNP |
DNP |
DNP |
Forward Compatibility |
DNP |
DNP |
DNP |
DNP |
DNP |
DNP |
DNP |
Of the candidates reviewed in this draft, two claim to satisfy all the minimum W3C
requirements. While any of the formats above could form the basis for a W3C standard,
it must be noted that those that satisfy fewer constraints would likely require
more modifications than those that already meet the minimum W3C requirements.
Each modification to a format will impact the processing efficiency of the format and
may also impact other characteristics, such as compactness. For example, it is possible (but not necessary)
that modifications required to meet the W3C compactness requirements may significantly
reduce processing efficiency and modifications required to improve processing efficiency
may impact compactness. As such, the performance characteristics of each
candidate presented in this table that does not meet all requirements illustrate what
is possible when certain W3C requirements are not met, but do not illustrate the
performance of a format based that candidate modified to comply to the W3C
requirements.
It must therefore be noted that analysis of candidates should take into account whether
candidates meet all or a subset of the W3C requirements.
3. Compaction Analysis
This section gives a preliminary analysis of the results from the current set of test groups
based on the latest runs performed with the candidates that have been
submitted.
3.1. W3C Compaction Metrics
Given the many ways in which compression can be achieved, and the fact that some ways
are appropriate for some usage scenarios but not for others, candidates are measured
in four compaction cases that cover this variety in usage.
- Neither
-
In this case the candidate has no access to external information such as a schema
or has it but the encoded instances that it produces are still self-contained,
and does not perform any compression bases on analysing
the document. Typically, simple tokenisation of the XML document is
performed.
- Document
-
In this case the candidate does not have access external information such as schema
or has it but the encoded instances that it produces are still self-contained,
but may perform various document-analysis operations such as
frequency-based compression (results are compared against gzipped XML).
- Schema
-
In this case the candidate does not perform any manner of document analysis but may
rely on externally provided information, typically a schema, and the resulting bitstream
may not be self-contained.
- Both
-
In this case the candidate may use methods available in both of the Document and the
Schema cases.
The following graphs provide a summary of the compactness results for each of
these compaction classes.
This analysis is based on the information density clusters (the number of
documents in the cluster is shown in parentheses). The nomenclature used
here is described in the EXI Measurements
Document section Characterization
of Compactness.
Note that the test documents do not divide neatly such that each test group
fits entirely into a single cluster, thereby entailing that documents from the
same test group frequently belong to separate information density clusters.
- High Information Density (16)
-
As can be expected, for the Neither case candidates mostly
stay above the information density of the documents. There are a few
exceptions though, in particular one of the GAML documents where all
candidates achieve a compactness of one half of the information
density. In the Document case, apart from Efficient XML the
candidates remain above 90% size in all cases except the GAML
document. Efficient XML manages to penetrate the 90% barrier four
times in total.
-
When allowed to use only Schema based compactification, both FXDI and Efficient XML manage to
consistently achieve sizes approximately 60% of XML, except for
SVGTiny documents where the sizes are between 80% and 90% and for
DataStore where the sizes are between 30% and 50%. Apart from the
pathological GAML document, the DataStore documents, and the Seismic
document, the ratio increases modestly in the Both case.
- Low Information Density - Large Documents (19)
-
Again in the Neither case the candidates remain above
the information density, with no anomalies this time. Of the
candidates, Efficient XML achieves consistently the best results, with
Fast Infoset a bit ahead of the rest. In the Document case, the
situation looks somewhat better than with the high density documents,
but still the 90% barrier seems a reasonable one. Compared to the
others, now Xebu is clearly behind the rest.
-
Only the densest of documents, in the OpenOffice group, go above the
25% mark with Schema information included but it must be noted that due
to schema issues the OpenOffice group results are not entirely consistent.
A few cases manage to penetrate even the 10% barrier. In the Both case, FXDI is almost
always between 80% and 100%, with Efficient XML managing to mostly go
between 70% and 80%.
- Low Information Density - Small Documents (33)
-
In the Neither case we do not see candidates managing
to achieve the information density of the documents, even Efficient
XML achieves a result above 50% more often than below, and the rest
manage to go below 50% only rarely. For the Document case, presumably
due to the small sizes, the situation compared to gzipped XML looks
better for the candidates, with FXDI and Efficient XML often managing
to go below 80%.
-
In the Schema case, apart from SVG, and XAL, going below 20% is more
the rule than an exception. Again, in a few cases the 10% barrier is
penetrated. With FixML the ratios are still clearly larger. The Both
case does not seem markedly different from the other clusters.
3.3. Analysis based on the use groups
- Scientific information
-
This use group shows very clearly that for instances that for strongly typed documents,
especially those that have a high information density (the archetype of which being the
Seismic test group), the benefits of being able to rely on schema information are high.
For the Seismic test group, candidates yield no compression in the Neither case, and
hit the same mark as gzip in the Document case. However in both cases in which schema
information is available candidates that use that information manage to reach 40% for
Schema, and 85% of gzip for Both (while candidates that don't fail to beat gzip).
-
Other test groups in this use group, given their lower ID, show much less caricatural
results. The dichotomy between candidates that make extensive use of schema information
(FXDI and Efficient XML) and those that make much less use of it (Xebu and Fast Infoset)
is still very noticeable in many cases but less pronounced. That having been said, despite
this use group constituting mostly of strongly typed information, schema information appears
to be by no means required for good compression levels, with Efficient XML almost breaking
the 5% barrier in the Neither case once, and all candidates except Xebu compressing to
50% of gzip in the MAGE-ML test group. In general the variance of results seems to be
a factor of the ability to use schema information and of the ID of individual test cases.
- Financial information
-
While the actual content of the test groups in this use group are related in many ways,
the FixML test group stands out in that its XML vocabulary was largely optimised for
size already, through extensive usage of attributes and sometimes radical shortening
of element and attribute names. This naturally tends to render its information density
higher than that of the other two test groups, and this translates directly in the results
to lesser compression. FpML is the most structured group (i.e. the one with the lowest
ID), and Invoice is closer to FixML with most of its instances as dense as the least
dense of the FixML instances. This has effects across all cases, but is most evident
in the Neither one where we see FixML in the 70-90% range, Invoice largely around 35-40%,
and FpML falling as low as 20% with Efficient XML. It is also visible in the Document
case in which candidates often fail to beat gzip for FixML, while they tend to revolve
around 80% of gzip for FpML, and succeed in dropping to 75% of gzip for Invoice (albeit
with strong disparity between the candidates which is worth further investigation).
-
When schema information is available a strong dichotomy between candidates that benefit
from it and those that don't surfaces. In the FixML test group, the former barely rise
above 35% and go down to 20% for Schema and hit in the 30-50% area for Both, while the
latter never do better than 55% and go as high as 95% for Schema and for Both do no
better than 85% of gzip and are sometimes over the 100% of gzip bar. The dichotomy is
less marked for FpML and Invoice with Fast Infoset tending to score in between Xebu
and the two others that are typically below 15% with Efficient XML breaking the 5%
barrier in one FpML instance.
- Electronic documents
-
In this use group schema results are unreliable since the OpenDocument schema is
currently broken while the SVG Tiny one is not available.
-
For OpenOffice, results tend to be rather regular: globally in the 30-40% range for
Neither (with Efficient XML dropping to 20% in one instance) and 80-90% of gzip for
Document (with Xebu being larger than gzip). For SVG, the test group is deliberately
distributed largely across the information density spectrum, and the results vary
accordingly, going down to 45% for Neither in some instances, but with data-intensive
documents giving the candidates the greatest difficulty in producing any compression
or in beating gzip. It is noteworthy however that cleaning up the SVG documents seems
to only benefit Efficient XML (in the Document case).
- Broadcast metadata
-
The schema rift is again visible in the Schema and Both cases, with schema-informed
candidates reaching as low as 10% in the Schema case while the others do no better than
45%.
-
It is interesting to note that in the Neither case candidates are mostly distributed
around 65%, but Efficient XML gets close to the 35% bar on one occasion. In the Document
case there is a notable difference between Xebu and Fast Infoset on one side that struggle to
do better than gzip — failing most of the time — while FXDI and Efficient XML on the other side are mostly
in the 80% area (with Efficient XML going to 60%). The latter discrepancy may be worth further
investigation since it is independent of the usage of schema information.
- Data storage
-
This use group is somewhat special in that its membership is a single test group. That
being said, that test group is somewhat representative of the usage consisting of storing
relational database information more or less directly in an XML document.
-
This use group exhibits a strong variance in information density, and that translates
directly to variance in the results for all cases. For the Neither case, in one instance
the best result is 45%, while for another it almost falls below 5%. Likewise in the Schema
case, with Efficient XML doing as good as 2% in one instance. The variance is attenuated
(but still clearly present) in the Document and Both
cases since compression dampens the
effects of excessive structure, but the candidates still succeed in falling as low as 50%
of gzip for Document, and 15% of gzip for Both.
3.4. Analysis based on the XBC compaction metrics in the various classes
Analysis for the Schema and Both case is not currently considered,
since ASN.1
PER numbers are not available and they form the basis of the
viability comparison. These analyses will be made as soon as those results
become available.
- Neither
-
All candidates achieve sufficient compactness with all the
test documents, except for Xebu with the FixML test group and all
candidates with the Seismic test group. This latter is caused by the
Seismic test document consisting of a list of floating-point values;
without document analysis or schema information there is little that a
format can do. A similar case is with some of the SVG test documents,
where attribute values contain long sequences of floating point
values, but here all candidates achieve a very small amount of
reduction.
- Document
-
Apart from a few isolated cases, candidate Xebu does not
achieve sufficient compactness throughout the test suite. The FixML
set proves problematic here as well: of the four candidates and five
documents, Efficient XML achieves sufficient compactness with three
and FXDI with one. These are the only cases in which Efficient XML does not hit the mark; FXDI
has with four others. Fast Infoset is in the middle of these
classes, missing the mark with approximately a third of the test documents.
3.5. Conclusion
It surfaces from this analysis that there are two major influences on the compaction
performance of candidates: information density, and limitations on the candidate's
ability to exploit schema information (either because none is available, or because
the candidate supports less of it). Neither of these is surprising. Since EXI
methods normally function by eliminating as much as possible of the structural
overhead, they can only work so well in cases in which there is little structure.
Likewise, the more information one has about an instance, the better one can
compress it.
It is interesting to note though that while in many cases the information density
appears to be the compression limit for the Neither case, it
is occasionally broken. Also, in the Document case candidates
often succeed in doing better than gzip, especially on the smaller documents.
Quite naturally, schema information being the means through which the overhead of the
lexical representation of typed values is revealed, it is the most evident mean of
breaking through the ID barrier.
4. Caveats
As the time of this report, the test suite and some of the candidates still had some
outstanding issues that should be taken into account to better understand the analysis.
Also, the candidates are not functionally homogenous.
These caveats do not prevent a useful perusal of the current results but should be
to be taken into account in order to better understand the analysis.
4.1. Schema-related Caveats
- Limited schema usage
-
Some candidates (notably Xebu and Fast Infoset) only make use
of schema information in a very limited fashion — largely restricted
to knowing the names of elements and attributes in advance – and
there is therefore discussion as to whether it is pertinent to
cover their analysis in the Schema and Both
cases below. (See Candidate-related Caveats.)
-
Xebu is able to make more extensive use of schema than creating a
table of names but it does not do so now. This is because it only
supports RelaxNG at this point and the schemata are not yet
available in both RelaxNG and XML Schema.
- Schema quality
-
Not all test groups include vocabularies for which complete
and normative schemata are available. Further analysis is
therefore desirable concerning the impact of using schema
or not.
- OpenDocument Schema
-
Due to technical issues with the OpenDocument schema, it is
currently not being used to produce the test results.
In the real world use case the schema would normally be used.
4.2. Candidate-related Caveats
- X.694 with PER
-
Complete results for this candidate have yet to be obtained. It must
be noted that this candidate does not have support for all of the
fidelity options, and that some of its results can therefore appear
smaller and faster than they would be if it did.
- X.694 with BER
-
Complete results for this candidate have yet to be obtained. It must
be noted that this candidate does not have support for all of the
fidelity options, and that some of its results can therefore appear
smaller and faster than they would be if it did.
- X.891 (Fast Infoset)
-
For the Schema and Both
application classes Fast Infoset is not fully optimised to utilise schema:
-
-
For the cases where prefixes do not need to be preserved (or in the
more general case when a set of sample documents can be utilised in
conjunction with schema) further reductions in compactness and
processing efficiency are possible.
-
The use of encoding algorithms or restricted alphabets to convert
lexical representations of text content or attribute values to more
optimised binary forms that may be faster to process and/or more
compact.
- Xebu
-
The current version is research software, therefore neither the format
nor the implementation can be considered to be fully mature.
- Efficient XML
-
The tests are being run using a non-production implementation of Efficient
XML designed for evaluating proposed extensions to Efficient XML. It includes
support for features beyond the minimum W3C requirements, such as random
access, accelerated sequential access and XML Digital Signatures. As a
non-production version, it has not been fully optimized and does not
achieve the same results as production releases of Efficient XML.
- XSBC
-
No caveats at this point.
- FXDI
-
No caveats at this point.
- esXML
-
This candidate has not yet been included in the test framework.
4.3. Test Group Caveats
- SVG Tiny, MAGE-ML, XAL
-
Some candidates are not yet able to preserve
DTDs; therefore, DTD
preservation has been temporarily disabled so as not to penalise the
compactness results of the candidates that can preserve them.
It will be turned back on as soon as these issues are properly addressed.
- SVG Tiny
-
The SVG Tiny test group also has the particularity of being present
twice: once with the original data and once more with a version
of the same information that has been cleaned up. The motivation
behind this is simple; many SVG authoring tools include all
manners of data in the documents they produce that are by and large
useless for the intended applications. Some SVG content publishers
will publish the output from these authoring tools directly, despite
the noticeable extra cost in bandwidth and processing time, while others
will add a clean-up process to the publication. Since many of the
documents in the original test group come from authoring tools that
produce "dirty" content, and given the fact that there are two ways
of publishing such content, the "clean" alternative has been added
so that as a whole the SVG data reflects actual patterns of
usage.
A. Test Groups Overview
This appendix describes the various test groups that are currently
in the test suite. As described in the Abstract,
the entirety of the test groups at our disposal has not yet been
included in these runs.
The test suite is organised into "test groups". Each group
consists of XML document instances pertaining to the same
vocabulary, or to a set of related applications of XML. For each
group, information is provided on the type of data that it
contains, and on the use cases from the
XBC Use Cases
document [XBC-UC] to
which it relates. Additionally, each group is annotated with its
fidelity level requirement, the lexical reproducibility
typically necessary for the group, from the table in
the Efficient XML
Interchange Measurements Note, section on
the Fidelity Scale.
5.2. BCAST
- Description
-
BAC Broadcast
is a vocabulary defined by the OMA
for use in the transport of broadcast metadata, notably
electronic service guides for mobile devices.
- Use cases
-
Metadata in Broadcast Systems
- Fidelity
-
Level -1. For the most part, only elements and attributes are required.
- Initial assessment
-
In progress.
5.3. CBMS
- Description
-
DVB
CBMS
(Convergence of Broadcast and Mobile Services, formerly known as UMTS) is a set of
standards aiming at improving convergence between 3G and broadcast services. The
samples in the test suite are used for electronic service guides.
- Use cases
-
Metadata in Broadcast Systems
- Fidelity
-
Level -1. For the most part, only elements and attributes are required.
- Initial assessment
-
In progress.
5.4. DataStore
- Description
-
This is a very simple format used to capture and store primitive data.
- Use cases
-
XML Documents in Persistent Store
- Fidelity
-
Level -1. Only relies on elements and attributes.
- Initial assessment
-
In progress.
5.8. Invoice
- Description
-
This set of documents was taken from transactions between companies and then
obfuscated. It represents typical invoice documents as exchanged between machines
both inside and outside companies.
- Use cases
-
Intra/Inter Business Communication
- Fidelity
-
Level -1. Essentially attributes and elements.
- Initial assessment
-
In progress.
5.9. MAGE-ML
- Description
-
The Microarray Gene Expression Markup Language (MAGE-ML)
is a language designed to describe and communicate information about microarray
based experiments. MAGE-ML is based on XML and can describe microarray designs,
microarray manufacturing information, microarray experiment setup and execution
information, gene expression data and data analysis results.
- Use cases
-
Supercomputing and Grid Processing
- Fidelity
-
Level 1. An Infoset level preservation is needed at least for the
Document Type Declaration.
- Initial assessment
-
In progress.
5.10. OpenOffice
- Description
-
The Open Document Format (ODF)
is a set of vocabularies for use within office applications.
- Use cases
-
Electronic Documents
- Fidelity
-
Level 0. Most constructs found in the XPath 1.0 data model are used
but none more.
- Initial assessment
-
In progress.
5.12. SVG Tiny
- Description
-
Scalable Vector Graphics (SVG)
is a language for describing high-quality two-dimensional, interactive,
animated, and scriptable graphics. It is increasingly available from Web
browsers, and its Tiny profile is the de facto standard for animated graphics
in the mobile industry.
- Use cases
-
Electronic Documents, Multimedia XML Documents for Mobile Handsets
- Fidelity
-
Level 2. It is not strictly necessary to preserve all of the
internal subset, but SVG is commonly hand-authored (or at least hand-edited after
the initial graphics have been created) and it is frequent that the internal
subset will be used to define entities for commonly used attribute values as well
as to declare some attributes to be of type
ID
.
- Initial assessment
-
In progress.
5.13. XAL
- Description
-
XAL is a software
framework for particle accelerator online modelling software. It
makes use of XML descriptions of the devices, such as magnets of
various field shapes, beam position monitors, and others,
and XML documents of the translation matrices which describe
how the beam propagates from one device to another.
- Use cases
-
Floating Point Arrays in the Energy Industry, Sensor Processing and Communication, Supercomputing and Grid Processing
- Fidelity
-
Level 1. An Infoset level preservation is needed at least for the
Document Type Declaration.
- Initial assessment
-
In progress.