Re: Recovery from decompression failure (was: Re: Cost analysis: (was: Getting to Consensus: CONTINUATION-related issues))

On 23 July 2014 17:55, Tatsuhiro Tsujikawa <tatsuhiro.t@gmail.com> wrote:
> So decoder changes table size only when it sees encodong context update? It
> works, but it has a disadvantage because decoder has to retain header table
> size unchanged until HEADERS/PP are received, which might never arrive.
> On the other hand, SETTINGS ACK is guaranteed to arrive (and we have a
> timeout for it). If decoder changes header table size, including eviction to
> the lowest value it advertised when it received SETTINGS ACK, encoder only
> sends its selected value based on the last seen value in SETTINGS decoder
> sent. This is because decoder already performed eviction based on the lowest
> value and encoding context update comes after SETTINGS ACK.

Yes, I would use SETTINGS ACK as the marker to use for trimming the
header table down, but still require the next header block after that
to include a context update.  That might mean two rounds of eviction
if the encoder chooses to use a smaller table size, but that's the
safest approach.

Received on Thursday, 24 July 2014 13:12:46 UTC