draft-ietf-quic-qcram-00.txt   draft-ietf-quic-qcram-latest.txt 
QUIC Working Group C. Krasic QUIC Working Group C. Krasic
Internet-Draft Google, Inc Internet-Draft Google, Inc
Intended status: Standards Track M. Bishop Intended status: Standards Track M. Bishop
Expires: August 24, 2018 Akamai Technologies Expires: January 7, 2025 Akamai Technologies
A. Frindell, Ed. A. Frindell, Ed.
Facebook Facebook
February 20, 2018 July 6, 2024
Header Compression for HTTP over QUIC Header Compression for HTTP over QUIC
draft-ietf-quic-qcram-00 draft-ietf-quic-qcram-latest
Abstract Abstract
The design of the core QUIC transport subsumes many HTTP/2 features, This specification defines QCRAM, a compression format for
prominent among them stream multiplexing. A key advantage of the efficiently representing HTTP header fields, to be used in HTTP over
QUIC transport is stream multiplexing free of head-of-line (HoL) QUIC. This is a variation of HPACK header compression that seeks to
blocking between streams. In HTTP/2, multiplexed streams can suffer reduce head-of-line blocking.
HoL blocking due to TCP.
If HTTP/2's HPACK is used for header compression, HTTP/QUIC is still Note to Readers
vulnerable to HoL blocking, because of HPACK's assumption of in-order
delivery. This draft defines QCRAM, a variation of HPACK and Discussion of this draft takes place on the QUIC working group
mechanisms in the HTTP/QUIC mapping that allow the flexibility to mailing list (quic@ietf.org), which is archived at
avoid header-compression-induced HoL blocking. https://mailarchive.ietf.org/arch/search/?email_list=quic [1].
Working Group information can be found at https://github.com/quicwg
[2]; source code and issues list for this draft can be found at
https://github.com/quicwg/base-drafts/labels/-qcram [3].
Status of This Memo Status of This Memo
This Internet-Draft is submitted in full conformance with the This Internet-Draft is submitted in full conformance with the
provisions of BCP 78 and BCP 79. provisions of BCP 78 and BCP 79.
Internet-Drafts are working documents of the Internet Engineering Internet-Drafts are working documents of the Internet Engineering
Task Force (IETF). Note that other groups may also distribute Task Force (IETF). Note that other groups may also distribute
working documents as Internet-Drafts. The list of current Internet- working documents as Internet-Drafts. The list of current Internet-
Drafts is at https://datatracker.ietf.org/drafts/current/. Drafts is at https://datatracker.ietf.org/drafts/current/.
Internet-Drafts are draft documents valid for a maximum of six months Internet-Drafts are draft documents valid for a maximum of six months
and may be updated, replaced, or obsoleted by other documents at any and may be updated, replaced, or obsoleted by other documents at any
time. It is inappropriate to use Internet-Drafts as reference time. It is inappropriate to use Internet-Drafts as reference
material or to cite them other than as "work in progress." material or to cite them other than as "work in progress."
This Internet-Draft will expire on August 24, 2018. This Internet-Draft will expire on January 7, 2025.
Copyright Notice Copyright Notice
Copyright (c) 2018 IETF Trust and the persons identified as the Copyright (c) 2024 IETF Trust and the persons identified as the
document authors. All rights reserved. document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal This document is subject to BCP 78 and the IETF Trust's Legal
Provisions Relating to IETF Documents Provisions Relating to IETF Documents
(https://trustee.ietf.org/license-info) in effect on the date of (https://trustee.ietf.org/license-info) in effect on the date of
publication of this document. Please review these documents publication of this document. Please review these documents
carefully, as they describe your rights and restrictions with respect carefully, as they describe your rights and restrictions with respect
to this document. Code Components extracted from this document must to this document. Code Components extracted from this document must
include Simplified BSD License text as described in Section 4.e of include Simplified BSD License text as described in Section 4.e of
the Trust Legal Provisions and are provided without warranty as the Trust Legal Provisions and are provided without warranty as
described in the Simplified BSD License. described in the Simplified BSD License.
Table of Contents Table of Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 2 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 2
1.1. Head-of-Line Blocking in HPACK . . . . . . . . . . . . . 3 1.1. Head-of-Line Blocking in HPACK . . . . . . . . . . . . . 3
1.2. Avoiding Head-of-Line Blocking in HTTP/QUIC . . . . . . . 3 1.2. Avoiding Head-of-Line Blocking in HTTP/QUIC . . . . . . . 3
2. HTTP over QUIC mapping extensions . . . . . . . . . . . . . . 4 2. HPACK extensions . . . . . . . . . . . . . . . . . . . . . . 4
2.1. HEADERS and PUSH_PROMISE . . . . . . . . . . . . . . . . 4 2.1. Allowed Instructions . . . . . . . . . . . . . . . . . . 4
2.2. HEADER_ACK . . . . . . . . . . . . . . . . . . . . . . . 5 2.2. Header Block Prefix . . . . . . . . . . . . . . . . . . . 5
3. HPACK extensions . . . . . . . . . . . . . . . . . . . . . . 5 2.3. Hybrid absolute-relative indexing . . . . . . . . . . . . 5
3.1. Allowed Instructions . . . . . . . . . . . . . . . . . . 5 2.4. Preventing Eviction Races . . . . . . . . . . . . . . . . 6
3.2. Header Block Prefix . . . . . . . . . . . . . . . . . . . 5 2.4.1. Blocked Evictions . . . . . . . . . . . . . . . . . . 7
3.3. Hybrid absolute-relative indexing . . . . . . . . . . . . 6 2.5. Refreshing Entries with Duplication . . . . . . . . . . . 7
3.4. Preventing Eviction Races . . . . . . . . . . . . . . . . 7 3. Performance considerations . . . . . . . . . . . . . . . . . 7
3.4.1. Blocked Evictions . . . . . . . . . . . . . . . . . . 7 3.1. Speculative table updates . . . . . . . . . . . . . . . . 7
3.5. Refreshing Entries with Duplication . . . . . . . . . . . 8 3.2. Additional state beyond HPACK. . . . . . . . . . . . . . 8
4. Performance considerations . . . . . . . . . . . . . . . . . 8 3.2.1. Vulnerable Entries . . . . . . . . . . . . . . . . . 8
4.1. Speculative table updates . . . . . . . . . . . . . . . . 8 3.2.2. Safe evictions . . . . . . . . . . . . . . . . . . . 8
4.2. Additional state beyond HPACK. . . . . . . . . . . . . . 8 3.2.3. Decoder Blocking . . . . . . . . . . . . . . . . . . 8
4.2.1. Vulnerable Entries . . . . . . . . . . . . . . . . . 8 3.2.4. Fixed overhead. . . . . . . . . . . . . . . . . . . . 9
4.2.2. Safe evictions . . . . . . . . . . . . . . . . . . . 9 4. Security Considerations . . . . . . . . . . . . . . . . . . . 9
4.2.3. Decoder Blocking . . . . . . . . . . . . . . . . . . 9 5. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 9
4.2.4. Fixed overhead. . . . . . . . . . . . . . . . . . . . 9 6. References . . . . . . . . . . . . . . . . . . . . . . . . . 9
5. Security Considerations . . . . . . . . . . . . . . . . . . . 10 6.1. Normative References . . . . . . . . . . . . . . . . . . 9
6. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 10 6.2. Informative References . . . . . . . . . . . . . . . . . 9
7. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . 10 6.3. URIs . . . . . . . . . . . . . . . . . . . . . . . . . . 10
8. References . . . . . . . . . . . . . . . . . . . . . . . . . 10 Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . 10
8.1. Normative References . . . . . . . . . . . . . . . . . . 10 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 10
8.2. Informative References . . . . . . . . . . . . . . . . . 11
Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 11
1. Introduction 1. Introduction
The QUIC transport protocol was designed from the outset to support The QUIC transport protocol was designed from the outset to support
HTTP semantics, and its design subsumes many of the features of HTTP semantics, and its design subsumes many of the features of
HTTP/2. QUIC's stream multiplexing comes into some conflict with HTTP/2. QUIC's stream multiplexing comes into some conflict with
header compression. A key goal of the design of QUIC is to improve header compression. A key goal of the design of QUIC is to improve
stream multiplexing relative to HTTP/2 by eliminating HoL (head of stream multiplexing relative to HTTP/2 by eliminating HoL (head of
line) blocking, which can occur in HTTP/2. HoL blocking can happen line) blocking, which can occur in HTTP/2. HoL blocking can happen
because all HTTP/2 streams are multiplexed onto a single TCP because all HTTP/2 streams are multiplexed onto a single TCP
skipping to change at page 4, line 8 skipping to change at page 4, line 10
1.2. Avoiding Head-of-Line Blocking in HTTP/QUIC 1.2. Avoiding Head-of-Line Blocking in HTTP/QUIC
In the example above, the second stream contained a reference to data In the example above, the second stream contained a reference to data
which might not yet have been processed by the recipient. Such which might not yet have been processed by the recipient. Such
references are called "vulnerable," because the loss of a different references are called "vulnerable," because the loss of a different
packet can keep the reference from being usable. packet can keep the reference from being usable.
The encoder can choose on a per-header-block basis whether to favor The encoder can choose on a per-header-block basis whether to favor
higher compression ratio (by permitting vulnerable references) or HoL higher compression ratio (by permitting vulnerable references) or HoL
resilience (by avoiding them). This is signaled by the BLOCKING flag resilience (by avoiding them). This is signaled by the BLOCKING flag
in HEADERS and PUSH_PROMISE frames (see Section 2). in HEADERS and PUSH_PROMISE frames (see [QUIC-HTTP]).
If a header block contains no vulnerable header fields, BLOCKING MUST If a header block contains no vulnerable header fields, BLOCKING MUST
be 0. This implies that the header fields are represented either as be 0. This implies that the header fields are represented either as
references to dynamic table entries which are known to have been references to dynamic table entries which are known to have been
received, or as Literal header fields (see [RFC7541] Section 6.2). received, or as Literal header fields (see Section 6.2 of [RFC7541]).
If a header block contains any header field which references dynamic If a header block contains any header field which references dynamic
table state which the peer might not have received yet, the BLOCKING table state which the peer might not have received yet, the BLOCKING
flag MUST be set. If the peer does not yet have the appropriate flag MUST be set. If the peer does not yet have the appropriate
state, such blocks might not be processed on arrival. state, such blocks might not be processed on arrival.
The header block contains a prefix (Section 3.2). This prefix The header block contains a prefix (Section 2.2). This prefix
contains table offset information that establishes total ordering contains table offset information that establishes total ordering
among all headers, regardless of reordering in the transport (see among all headers, regardless of reordering in the transport (see
Section 3.3). Section 2.3).
In blocking mode, the prefix additionally identifies the minimum In blocking mode, the prefix additionally identifies the minimum
state required to process any vulnerable references in the header state required to process any vulnerable references in the header
block (see "Depends Index" in Section 3.3). The decoder keeps track block (see "Depends Index" in Section 2.3). The decoder keeps track
of which entries have been added to its dynamic table. The stream of which entries have been added to its dynamic table. The stream
for a header with BLOCKING flag set is considered blocked by the for a header with BLOCKING flag set is considered blocked by the
decoder and can not be processed until all entries in the range "[1, decoder and can not be processed until all entries in the range "[1,
Depends Index]" have been added. While blocked, header field data Depends Index]" have been added. While blocked, header field data
MUST remain in the blocked stream's flow control window. MUST remain in the blocked stream's flow control window.
2. HTTP over QUIC mapping extensions 2. HPACK extensions
2.1. HEADERS and PUSH_PROMISE
HEADERS and PUSH_PROMISE frames define a new flag.
BLOCKING (0x01): Indicates the stream might need to wait for
dependent headers before processing. If 0, the frame can be
processed immediately upon receipt.
HEADERS frames can be sent on the Connection Control Stream as well
as on request / push streams. The value of BLOCKING MUST be 0 for
HEADERS frames on the Connection Control Stream, since they can only
depend on previous HEADERS on the same stream.
2.2. HEADER_ACK
The HEADER_ACK frame (type=0x8) is sent from the decoder to the
encoder on the Control Stream when the decoder has fully processed a
header block. It is used by the encoder to determine whether
subsequent indexed representations that might reference that block
are vulnerable to HoL blocking, and to prevent eviction races (see
Section 3.4).
The HEADER_ACK frame indicates the stream on which the header block
was processed by encoding the Stream ID as a variable-length integer.
The same Stream ID can be identified multiple times, as multiple
header-containing blocks can be sent on a single stream in the case
of intermediate responses, trailers, pushed requests, etc. as well as
on the Control Streams. Since header frames on each stream are
received and processed in order, this gives the encoder precise
feedback on which header blocks within a stream have been fully
processed.
0 1 2 3 4 5 6 7
+---+---+---+---+---+---+---+---+
| Stream ID [i] |
+---+---------------------------+
HEADER_ACK frame
The HEADER_ACK frame does not define any flags.
3. HPACK extensions
3.1. Allowed Instructions 2.1. Allowed Instructions
HEADERS frames on the Control Stream SHOULD contain only Literal with HEADERS frames on the Control Stream SHOULD contain only Literal with
Incremental Indexing and Indexed with Duplication (see Section 3.5) Incremental Indexing and Indexed with Duplication (see Section 2.5)
representations. Frames on this stream modify the dynamic table representations. Frames on this stream modify the dynamic table
state without generating output to any particular request. state without generating output to any particular request.
HEADERS and PUSH_PROMISE frames on request and push streams MUST NOT HEADERS and PUSH_PROMISE frames on request and push streams MUST NOT
contain Literal with Incremental Indexing and Indexed with contain Literal with Incremental Indexing and Indexed with
Duplication representations. Frames on these streams reference the Duplication representations. Frames on these streams reference the
dynamic table in a particular state without modifying it, but emit dynamic table in a particular state without modifying it, but emit
the headers for an HTTP request or response. the headers for an HTTP request or response.
3.2. Header Block Prefix 2.2. Header Block Prefix
For request and push promise streams, in HEADERS and PUSH_PROMISE For request and push promise streams, in HEADERS and PUSH_PROMISE
frames, HPACK Header data is prefixed by an integer: "Base Index". frames, HPACK Header data is prefixed by an integer: "Base Index".
"Base index" is the cumulative number of entries added to the dynamic "Base index" is the cumulative number of entries added to the dynamic
table prior to encoding the current block, including any entries table prior to encoding the current block, including any entries
already evicted. It is encoded as a single 8-bit prefix integer: already evicted. It is encoded as a single 8-bit prefix integer:
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7
+-+-+-+-+-+-+-+-+ +-+-+-+-+-+-+-+-+
|Base Index (8+)| |Base Index (8+)|
+---------------+ +---------------+
Figure 1: Absolute indexing (BLOCKING=0x0) Figure 1: Absolute indexing (BLOCKING=0x0)
Section 3.3 describes the role of "Base Index". Section 2.3 describes the role of "Base Index".
When the BLOCKING flag is 0x1, a the prefix additionally contains a When the BLOCKING flag is 0x1, a the prefix additionally contains a
second HPACK integer (8-bit prefix) 'Depends': second HPACK integer (8-bit prefix) 'Depends':
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7
+-+-+-+-+-+-+-+-+ +-+-+-+-+-+-+-+-+
|Base Index (8+)| |Base Index (8+)|
+---------------+ +---------------+
|Depends (8+)| |Depends (8+)|
+---------------+ +---------------+
Figure 2: Absolute indexing (BLOCKING=0x1) Figure 2: Absolute indexing (BLOCKING=0x1)
Depends is used to identify header dependencies (see Section 1.2). Depends is used to identify header dependencies (see Section 1.2).
The encoder computes a value "Depends Index" which is the largest The encoder computes a value "Depends Index" which is the largest
(absolute) index referenced by the following header block. To help (absolute) index referenced by the following header block. To help
keep the prefix smaller, "Depends Index" is converted to a relative keep the prefix smaller, "Depends Index" is converted to a relative
value: "Depends = Base Index - Depends Index". value: "Depends = Base Index - Depends Index".
3.3. Hybrid absolute-relative indexing 2.3. Hybrid absolute-relative indexing
HPACK indexed entries refer to an entry by its current position in HPACK indexed entries refer to an entry by its current position in
the dynamic table. As Figure 1 of [RFC7541] illustrates, newer the dynamic table. As Figure 1 of [RFC7541] illustrates, newer
entries have smaller indices, and older entries are evicted first if entries have smaller indices, and older entries are evicted first if
the table is full. Under this scheme, each insertion to the table the table is full. Under this scheme, each insertion to the table
causes the index of all existing entries to change (implicitly). causes the index of all existing entries to change (implicitly).
Implicit index updates are acceptable for HTTP/2 because TCP is Implicit index updates are acceptable for HTTP/2 because TCP is
totally ordered, but are problematic in the out-of-order context of totally ordered, but are problematic in the out-of-order context of
QUIC. QUIC.
skipping to change at page 7, line 4 skipping to change at page 6, line 9
Implicit index updates are acceptable for HTTP/2 because TCP is Implicit index updates are acceptable for HTTP/2 because TCP is
totally ordered, but are problematic in the out-of-order context of totally ordered, but are problematic in the out-of-order context of
QUIC. QUIC.
QCRAM uses a hybrid absolute-relative indexing approach. QCRAM uses a hybrid absolute-relative indexing approach.
When the encoder adds a new entry to its header table, it can compute When the encoder adds a new entry to its header table, it can compute
an absolute index: an absolute index:
"entry.absoluteIndex = baseIndex++;" "entry.absoluteIndex = baseIndex++;"
Since literals with indexing are only sent on the control stream, the Since literals with indexing are only sent on the control stream, the
decoder can be guaranteed to compute the same absolute index values decoder can be guaranteed to compute the same absolute index values
when it adds corresponding entries to its table, just as in HPACK and when it adds corresponding entries to its table, just as in HPACK and
HTTP/2. HTTP/2.
When encoding indexed representations, the following holds for When encoding indexed representations, the following holds for
(relative) HPACK indices: (relative) HPACK indices:
"relative index = baseIndex - entry.absoluteIndex + staticTable.size" "relative index = baseIndex - entry.absoluteIndex + staticTable.size"
Header blocks on request and push streams do not modify the dynamic Header blocks on request and push streams do not modify the dynamic
table state, so they never change the "baseIndex". However, since table state, so they never change the "baseIndex". However, since
ordering between streams is not guaranteed, the value of "baseIndex" ordering between streams is not guaranteed, the value of "baseIndex"
can not be synchronized implicitly. Instead then, QCRAM sends can not be synchronized implicitly. Instead then, QCRAM sends
encoder's "Base Index" explicitly as part of the prefix (see encoder's "Base Index" explicitly as part of the prefix (see
Section 3.2), so that the decoder can compute the same absolute Section 2.2), so that the decoder can compute the same absolute
indices that the encoder used: indices that the encoder used:
"absoluteIndex = prefix.baseIndex + staticTable.size - "absoluteIndex = prefix.baseIndex + staticTable.size -
relativeIndex;" relativeIndex;"
In this way, even if request or push stream headers are decoded in a In this way, even if request or push stream headers are decoded in a
different order than encoded, the absolute indices will still different order than encoded, the absolute indices will still
identify the correct table entries. identify the correct table entries.
It is an error if the HPACK decoder encounters an indexed It is an error if the HPACK decoder encounters an indexed
representation that refers to an entry missing from the table, and representation that refers to an entry missing from the table, and
the connection MUST be closed with the the connection MUST be closed with the
"HTTP_HPACK_DECOMPRESSION_FAILED" error code. "HTTP_HPACK_DECOMPRESSION_FAILED" error code.
3.4. Preventing Eviction Races 2.4. Preventing Eviction Races
Due to out-of-order arrival, QCRAM's eviction algorithm requires Due to out-of-order arrival, QCRAM's eviction algorithm requires
changes (relative to HPACK) to avoid the possibility that an indexed changes (relative to HPACK) to avoid the possibility that an indexed
representation is decoded after the referenced entry has already been representation is decoded after the referenced entry has already been
evicted. QCRAM employs a two-phase eviction algorithm, in which the evicted. QCRAM employs a two-phase eviction algorithm, in which the
encoder will not evict entries that have outstanding (unacknowledged) encoder will not evict entries that have outstanding (unacknowledged)
references. references.
3.4.1. Blocked Evictions 2.4.1. Blocked Evictions
The encoder MUST NOT permit an entry to be evicted while a reference The encoder MUST NOT permit an entry to be evicted while a reference
to that entry remains unacknowledged. If a new header to be inserted to that entry remains unacknowledged. If a new header to be inserted
into the dynamic table would cause the eviction of such an entry, the into the dynamic table would cause the eviction of such an entry, the
encoder MUST NOT emit the insert instruction until the reference has encoder MUST NOT emit the insert instruction until the reference has
been processed by the decoder and acknowledged. been processed by the decoder and acknowledged.
The encoder can emit a literal representation for the new header in The encoder can emit a literal representation for the new header in
order to avoid encoding delays, and MAY insert the header into the order to avoid encoding delays, and MAY insert the header into the
table later if desired. table later if desired.
To ensure that the blocked eviction case is rare, references to the To ensure that the blocked eviction case is rare, references to the
oldest entries in the dynamic table SHOULD be avoided. When one of oldest entries in the dynamic table SHOULD be avoided. When one of
the oldest entries in the table is still actively used for the oldest entries in the table is still actively used for
references, the encoder SHOULD emit an Indexed-Duplicate references, the encoder SHOULD emit an Indexed-Duplicate
representation instead (see Section 3.5). representation instead (see Section 2.5).
3.5. Refreshing Entries with Duplication 2.5. Refreshing Entries with Duplication
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7
+-+-+-+-+-+-+-+-+ +-+-+-+-+-+-+-+-+
|0|0|1|Index(5+)| |0|0|1|Index(5+)|
+-+-+-+---------+ +-+-+-+---------+
Figure 3: Indexed Header Field with Duplication Figure 3: Indexed Header Field with Duplication
_Indexed-Duplicates_ insert a new entry into the dynamic table which _Indexed-Duplicates_ insert a new entry into the dynamic table which
duplicates an existing entry. [RFC7541] allows duplicate HPACK table duplicates an existing entry. [RFC7541] allows duplicate HPACK table
entries, that is entries that have the same name and value. entries, that is entries that have the same name and value.
This replaces the HPACK instruction for Dynamic Table Size Update This replaces the HPACK instruction for Dynamic Table Size Update
(see Section 6.3 of [RFC7541], which is not supported by HTTP over (see Section 6.3 of [RFC7541], which is not supported by HTTP over
QUIC. QUIC.
4. Performance considerations 3. Performance considerations
4.1. Speculative table updates 3.1. Speculative table updates
Implementations can _speculatively_ send header frames on the HTTP Implementations can _speculatively_ send header frames on the HTTP
Control Streams which are not needed for any current HTTP request or Control Streams which are not needed for any current HTTP request or
response. Such headers could be used strategically to improve response. Such headers could be used strategically to improve
performance. For instance, the encoder might decide to _refresh_ by performance. For instance, the encoder might decide to _refresh_ by
sending Indexed-Duplicate representations for popular header fields sending Indexed-Duplicate representations for popular header fields
(Section 3.2), ensuring they have small indices and hence minimal (Section 2.2), ensuring they have small indices and hence minimal
size on the wire. size on the wire.
4.2. Additional state beyond HPACK. 3.2. Additional state beyond HPACK.
4.2.1. Vulnerable Entries 3.2.1. Vulnerable Entries
For header blocks encoded in non-blocking mode, the encoder needs to For header blocks encoded in non-blocking mode, the encoder needs to
forego indexed representations that refer to vulnerable entries (see forego indexed representations that refer to vulnerable entries (see
Section 1.2). An implementation could extend the header table entry Section 1.2). An implementation could extend the header table entry
with a boolean to track vulnerability. However, the number of with a boolean to track vulnerability. However, the number of
entries in the table that are vulnerable is likely to be small in entries in the table that are vulnerable is likely to be small in
practice, much less than the total number of entries, so a data practice, much less than the total number of entries, so a data
tracking only vulnerable (un-acknowledged) entries, separate from the tracking only vulnerable (un-acknowledged) entries, separate from the
main header table, might be more space efficient. main header table, might be more space efficient.
4.2.2. Safe evictions 3.2.2. Safe evictions
Section Section 3.4 describes how QCRAM avoids invalid references Section Section 2.4 describes how QCRAM avoids invalid references
that might result from out-of-order delivery. When the encoder that might result from out-of-order delivery. When the encoder
processes a HEADER_ACK, it dereferences table entries that were processes a HEADER_ACK, it dereferences table entries that were
indexed in the acknowledged header. To track which entries must be indexed in the acknowledged header. To track which entries must be
dereferenced, it can maintain a map from unacknowledged headers to dereferenced, it can maintain a map from unacknowledged headers to
lists of (absolute) indices. The simplest place to store the actual lists of (absolute) indices. The simplest place to store the actual
reference count might be the table entries. In practice the number reference count might be the table entries. In practice the number
of entries in the table with a non-zero reference count is likely to of entries in the table with a non-zero reference count is likely to
stay quite small. A data structure tracking only entries with non- stay quite small. A data structure tracking only entries with non-
zero reference counts, separate from the main header table, could be zero reference counts, separate from the main header table, could be
more space efficient. more space efficient.
4.2.3. Decoder Blocking 3.2.3. Decoder Blocking
To support blocking, the decoder needs to keep track of entries it To support blocking, the decoder needs to keep track of entries it
has added to the dynamic table (see Section 1.2), and it needs to has added to the dynamic table (see Section 1.2), and it needs to
track blocked streams. track blocked streams.
Tracking added entries might be done in a brute force fashion without Tracking added entries might be done in a brute force fashion without
additional space. However, this would have O(N) cost where N is the additional space. However, this would have O(N) cost where N is the
number of entries in the dynamic table. Alternatively, a dedicated number of entries in the dynamic table. Alternatively, a dedicated
data structure might improve on brute force in exchange a small data structure might improve on brute force in exchange a small
amount of additional space. For example, a set of pairs (of amount of additional space. For example, a set of pairs (of
indices), representing non-overlapping sub-ranges can be used. Each indices), representing non-overlapping sub-ranges can be used. Each
operation (add, or query) can be done within O(log M) complexity. operation (add, or query) can be done within O(log M) complexity.
Here set size M is the number of sub-ranges. In practice M would be Here set size M is the number of sub-ranges. In practice M would be
very small, as most table entries would be concentrated in the first very small, as most table entries would be concentrated in the first
sub-range [1, M]. sub-range "[1,M]".
To track blocked streams, an ordered map (e.g. multi-map) from To track blocked streams, an ordered map (e.g. multi-map) from
"Depends Index" values to streams can be used. Whenever the decoder "Depends Index" values to streams can be used. Whenever the decoder
processes a header block, it can drain any members of the blocked processes a header block, it can drain any members of the blocked
streams map that have "Depends Index <= M" where "[1,M]" is the first streams map that have "Depends Index <= M" where "[1,M]" is the first
member of the added- entries sub-ranges set. Again, the complexity member of the added- entries sub-ranges set. Again, the complexity
of operations would be at most O(log N), N being the number of of operations would be at most O(log N), N being the number of
concurrently blocked streams. concurrently blocked streams.
4.2.4. Fixed overhead. 3.2.4. Fixed overhead.
HPACK defines overhead as 32 bytes ([RFC7541] Section 4.1). As HPACK defines overhead as 32 bytes ([RFC7541], Section 4.1). As
described above, QCRAM adds some per-connection state, and possibly described above, QCRAM adds some per-connection state, and possibly
some per-entry state to track acknowledgment status and eviction some per-entry state to track acknowledgment status and eviction
reference count. A larger value than 32 might be more accurate for reference count. A larger value than 32 might be more accurate for
QCRAM. QCRAM.
5. Security Considerations 4. Security Considerations
TBD. TBD.
6. IANA Considerations 5. IANA Considerations
This document registers a new frame type, HEADER_ACK, for HTTP/QUIC. This document registers a new frame type, HEADER_ACK, for HTTP/QUIC.
This will need to be added to the IANA Considerations of [QUIC-HTTP]. This will need to be added to the IANA Considerations of [QUIC-HTTP].
7. Acknowledgments 6. References
This draft draws heavily on the text of [RFC7541]. The indirect
input of those authors is gratefully acknowledged, as well as ideas
from:
o Mike Bishop
o Alan Frindell
o Ryan Hamilton
o Patrick McManus
o Kazuho Oku
o Biren Roy
o Ian Swett
o Dmitri Tikhonov
8. References
8.1. Normative References 6.1. Normative References
[QUIC-HTTP] [QUIC-HTTP]
Bishop, M., "Hypertext Transfer Protocol (HTTP) over Bishop, M., "HTTP/3", draft-ietf-quic-http-34 (work in
QUIC", draft-ietf-quic-http-09 (work in progress), January progress), February 2021.
2018.
[RFC7541] Peon, R. and H. Ruellan, "HPACK: Header Compression for [RFC7541] Peon, R. and H. Ruellan, "HPACK: Header Compression for
HTTP/2", RFC 7541, DOI 10.17487/RFC7541, May 2015, HTTP/2", RFC 7541, DOI 10.17487/RFC7541, May 2015,
<https://www.rfc-editor.org/info/rfc7541>. <https://www.rfc-editor.org/info/rfc7541>.
8.2. Informative References 6.2. Informative References
[QUIC-TRANSPORT] [QUIC-TRANSPORT]
Iyengar, J. and M. Thomson, "QUIC: A UDP-Based Multiplexed Iyengar, J. and M. Thomson, "QUIC: A UDP-Based Multiplexed
and Secure Transport", draft-ietf-quic-transport-09 (work and Secure Transport", draft-ietf-quic-transport-34 (work
in progress), January 2018. in progress), January 2021.
[RFC7540] Belshe, M., Peon, R., and M. Thomson, Ed., "Hypertext [RFC7540] Belshe, M., Peon, R., and M. Thomson, Ed., "Hypertext
Transfer Protocol Version 2 (HTTP/2)", RFC 7540, Transfer Protocol Version 2 (HTTP/2)", RFC 7540,
DOI 10.17487/RFC7540, May 2015, DOI 10.17487/RFC7540, May 2015,
<https://www.rfc-editor.org/info/rfc7540>. <https://www.rfc-editor.org/info/rfc7540>.
6.3. URIs
[1] https://mailarchive.ietf.org/arch/search/?email_list=quic
[2] https://github.com/quicwg
[3] https://github.com/quicwg/base-drafts/labels/-qcram
Acknowledgments
This draft draws heavily on the text of [RFC7541]. The indirect
input of those authors is gratefully acknowledged, as well as ideas
from:
o Ryan Hamilton
o Patrick McManus
o Kazuho Oku
o Biren Roy
o Ian Swett
o Dmitri Tikhonov
Authors' Addresses Authors' Addresses
Charles 'Buck' Krasic Charles 'Buck' Krasic
Google, Inc Google, Inc
Email: ckrasic@google.com Email: ckrasic@google.com
Mike Bishop Mike Bishop
Akamai Technologies Akamai Technologies
 End of changes. 44 change blocks. 
141 lines changed or deleted 103 lines changed or added

This html diff was produced by rfcdiff 1.48. The latest version is available from http://tools.ietf.org/tools/rfcdiff/