Re: [quicwg/base-drafts] VarInt all the HTTP things (#2437)

[Lucas Pardue <[e-mail]>]

LPardue requested changes on this pull request.

I requested some changes, but if they are controversial we can move them to separate issues for further discussion.

> @@ -303,15 +303,16 @@ specify a value of zero for the QUIC transport parameter
 ## Unidirectional Streams
 
 Unidirectional streams, in either direction, are used for a range of purposes.
-The purpose is indicated by a stream type, which is sent as a single byte header
-at the start of the stream. The format and structure of data that follows this
-header is determined by the stream type.
+The purpose is indicated by a stream type, which is sent as a variable-length
+integer at the start of the stream. The format and structure of data that

you've replaced 'byte header' with 'variable-length integer', but the next sentence refers to 'this header'. This might be confusing to the reader? 

> @@ -373,7 +374,7 @@ MUST be treated as a stream error of type HTTP_WRONG_STREAM_DIRECTION.
  0                   1                   2                   3
  0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
-|Stream Type (8)|                  Push ID (i)                ...
+|    0x01 (8)   |                  Push ID (i)                ...

The alternative option is to remove it from the figure altogether. It was not great before the change - see #2338 

> @@ -433,12 +437,10 @@ All frames have the following format:
 A frame includes the following fields:
 
   Type:
-  : An 8-bit type for the frame.
-
+  : A variable-length type for the frame.

nit: make this consistent with the frame payload description. e.g.

A variable-length integer that identifies the frame type.

> @@ -671,14 +673,16 @@ treat the presence of the same parameter more than once as a connection error of
 type HTTP_MALFORMED_FRAME.
 
 The payload of a SETTINGS frame consists of zero or more parameters, each
-consisting of an unsigned 16-bit setting identifier and a value which uses the
+consisting of a setting identifier and a value, each of which uses the

"both encoded as" would work too
 

> @@ -881,12 +885,12 @@ the DUPLICATE_PUSH.
 
 ### Reserved Frame Types {#frame-grease}
 
-Frame types of the format `0xb + (0x1f * N)` are reserved to exercise the
-requirement that unknown types be ignored ({{extensions}}). These frames have no
-semantic value, and can be sent when application-layer padding is desired. They
-MAY also be sent on connections where no request data is currently being
-transferred. Endpoints MUST NOT consider these frames to have any meaning upon
-receipt.
+Frame types of the format `0x1f * N` for non-zero values of N are reserved to
+exercise the requirement that unknown types be ignored ({{extensions}}). These
+frames have no semantic value, and can be sent when application-layer padding is
+desired. They MAY also be sent on connections where no request data is currently

Not a new problem but I wonder if request data can be simplified to data, which covers all cases such as QPACK instructions, response data, control data etc.

> @@ -1589,15 +1593,9 @@ The entries in the following table are registered by this document.
 | NUM_PLACEHOLDERS             |  0x8   | {{settings-parameters}}   |
 | ---------------------------- | ------ | ------------------------- |

Not a new issue but: RFC 7540 has a column for `Initial Value`. I think it makes sense that HTTP/3 does similar.

> @@ -1868,7 +1861,9 @@ use the full 32-bit space.  Settings ported from HTTP/2 might choose to redefine
 the format of their settings to avoid using the 62-bit encoding.
 
 Settings need to be defined separately for HTTP/2 and HTTP/3. The IDs of
-settings defined in {{!RFC7540}} have been reserved for simplicity. See
+settings defined in {{!RFC7540}} have been reserved for simplicity.  Note that
+the settings identifier space in HTTP/3 is substantially larger (62 bits versus
+16 bits), so many HTTP/3 settings have no equivalent HTTP/2 code point. See

Unfortunately, `0x1` is reserved by QPACK so it makes a little more difficult to see that this is true. And a quick check of the [HTTP/2 IANA registry](https://www.iana.org/assignments/http2-parameters/http2-parameters.xhtml#settings) shows that 0x0 is also reserved, but I can't seem to find where that is defined in RFC 7540.

Checking that, it seems we have a clash on 0x8 due to RFC 8441 - WebSocket over HTTP/2. I think its reasonable that someone might write a WebSocket over HTTP/3 extension and reusing the code point would be nice. Can we move NUM_PLACEHOLDERS to 0x9 and leave 0x8 open?

-- 
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/quicwg/base-drafts/pull/2437#pullrequestreview-201516984