Stateless Reset Cleanup #1328
Stateless Reset Cleanup #1328
Conversation
The Stateless Reset is just a bunch of random bytes except the first one and the token, so I simplified the spec to reflect that.
Improved logical ordering of paragraphs.
|
The spec is clear that the CID is random, so it's all just random bytes between the short header byte and the token. However, an alternate way to improve this section is to leave it pretty much the same, except to say that if the stateless reset is in response to a long header, servers should use the SCID instead. Either is fine with me. |
| length, followed by a Stateless Reset Token. | ||
|
|
||
| The server SHOULD send a packet with a short header and a packet number length | ||
| of 1 octet. Using the hortest possible packet number encoding minimizes the |
| packet number length implied by the type field. A packet of this length | ||
| simulates a packet with a short header, the maximum-length destination | ||
| connection ID, a packet number of the appropriate length and (with the | ||
| Stateless Reset Token) an authentication trailer, with no packet payload. |
There was a problem hiding this comment.
Are fully-empty packets permitted? If not, perhaps we should mandate that this be slightly longer, so it appears there's a payload.
There was a problem hiding this comment.
Yes, let's not leak that. The simple fix here is s/18/19/
| Stateless Reset Token) an authentication trailer, with no packet payload. | ||
|
|
||
| In response to a packet with a short header, the server has no means of | ||
| determining the source connection ID and is effectively using a random one |
There was a problem hiding this comment.
It might be worth being precise about source in which direction. Maybe "The server has no means of determining a short packet's source connection ID, and therefore uses a random destination ID in the Stateless Reset packet."
There was a problem hiding this comment.
Reasonable, though make sure to state "destination connection ID"
MikeBishop
added
editorial
There was a problem hiding this comment.
This is a fine change, though I have a few comments. (If @martinduke doesn't get around to these, I will do so once I've finished my post-break triage.)
| packet number length implied by the type field. A packet of this length | ||
| simulates a packet with a short header, the maximum-length destination | ||
| connection ID, a packet number of the appropriate length and (with the | ||
| Stateless Reset Token) an authentication trailer, with no packet payload. |
There was a problem hiding this comment.
Yes, let's not leak that. The simple fix here is s/18/19/
| Stateless Reset Token) an authentication trailer, with no packet payload. | ||
|
|
||
| In response to a packet with a short header, the server has no means of | ||
| determining the source connection ID and is effectively using a random one |
There was a problem hiding this comment.
Reasonable, though make sure to state "destination connection ID"
| In response to a packet with a short header, the server has no means of | ||
| determining the source connection ID and is effectively using a random one | ||
| instead. For a client that depends on the server including a connection ID, | ||
| this will mean that this value differs from previous packets. Ths results in |
|
OBE |
The Stateless Reset is just a bunch of random bytes except the first one and the token, so I simplified the spec to reflect that.