RE: Position of HPACK Dynamic Table Size Update

I can’t speak to everyone’s implementation, but I don’t believe there’s a restriction on where in the packet an encoder can choose to change the table size.  We discussed, for example, the ability to clear the table by issuing back-to-back changes to zero and back to non-zero.

I would restate the highlighted text in the table as “if the SETTINGS_MAX_TABLE_SIZE changes such that the current size is no longer valid, the encoder MUST emit a table size change instruction prior to emitting any new compressed header blocks.”

From: Tatsuhiro Tsujikawa [mailto:tatsuhiro.t@gmail.com]
Sent: Wednesday, March 29, 2017 8:33 AM
To: ietf-http-wg@w3.org
Subject: Position of HPACK Dynamic Table Size Update

Hi,

I'd like to clarify the requirement of the position of HPACK Dynamic Table Size Update.
Should it appear always at the beginning of the header block?  Or is it only required to do so only if it is the ack of SETTINGS?
Background: https://github.com/summerwind/h2spec/issues/78

Best regards,
Tatsuhiro Tsujikawa