[quicwg/base-drafts] 3+ and 4+ in the QPACK wire format are complex and inefficient (#3561)

[Kazu Yamamoto <[e-mail]>]

For instance, QPACK uses the following wire format:

```
  0   1   2   3   4   5   6   7
+---+---+---+---+---+---+---+---+
| 0 | 0 | 1 | N | H |NameLen(3+)|
+---+---+---+---+---+-----------+
|  Name String (Length bytes)   |
+---+---------------------------+
| H |     Value Length (7+)     |
+---+---------------------------+
|  Value String (Length bytes)  |
+-------------------------------+
```

The designers of QPACK make use of 3 bits of the flag part. But HPACK uses the following wire format:

```
0   1   2   3   4   5   6   7
+---+---+---+---+---+---+---+---+
| 0 | 0 | 0 | 0 |       0       |
+---+---+-----------------------+
| H |     Name Length (7+)      |
+---+---------------------------+
|  Name String (Length octets)  |
+---+---------------------------+
| H |     Value Length (7+)     |
+---+---------------------------+
| Value String (Length octets)  |
+-------------------------------+
```

The designers of HPACK did not use 4 bits and chose 7+ for name length.

Let's calculate integer ranges of one/two/three bytes for each N+:

```
N+   1   2     3 <- bytes
8  127 382 16638
7   63 254 16510
6   31 190 16446
5   15 158 16414
4    7 142 16398
3    3 134 16390
2    1 130 16386
1    0 128 16384
```

One byte of 3+ can only express 0 - 3. In most cases, two bytes of 3+ are necessary for the length of header names. 

So, why don't we use 7+ for simplicity?

```
  0   1   2   3   4   5   6   7
+---+---+---+---+---+---+---+---+
| 0 | 0 | 1 | N | H |     0     |
+---+---+---+---+---+-----------+
| H |     Name Length (7+)      |
+---+---------------------------+
|  Name String (Length bytes)   |
+---+---------------------------+
| H |     Value Length (7+)     |
+---+---------------------------+
|  Value String (Length bytes)  |
+-------------------------------+
```

This keeps the length of wire format in most cases and enables us to reuse the code of HPACK.


-- 
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/issues/3561