Re: Call for Consensus: Remove "reference set" from HPACK (to address #552)

On Wed, Jul 16, 2014 at 04:00:33PM +1000, Greg Wilkins wrote:
> I don't think it is correct to say for 'almost all literal lookups'.   Any
> field with a custom value (with or without an indexed name) will be put
> into the header table.  The first 64 of those (ie the last 64) will be
> indexed with a single byte.   You only need go to a 2 byte index if the
> header table is large enough to hold more than 64 entries and your entry
> was added more that 64 additions ago.    I think 2 byte indexes are going
> to be rare.

I mean, the current design allows you to have a header field whose value
changes between consecutive requests (eg: a proxy aggregating multiple
users' requests), and still have that changing value referenced with
little overhead. Here to achieve the same result we'll need to have
that header field present in the static table in order for the index to
fit in a 6-bit quantity. I *think* that we have most common fields in the
static table so that should not be an issue. But that's something we might
want to double-check. Note that I'm clearly speaking about changing a
header value, not reusing the same value (which needs only one byte for
all cases).

Willy

Received on Wednesday, 16 July 2014 06:08:48 UTC