Re: Rechartering HTTPbis

On Thu, 2012-01-26 at 11:30 +0100, Willy Tarreau wrote:
> 
> Not everywhere. From what I observe at various places, what takes a lot
> of space in requests is :
>   - user agent : no need to compress it, just specify a new non-abusive format

I'll just say that there are an enormous number of voices that want to
influence the user-agent header. I don't believe any statement from the
HTTP-WG would have any impact other than making the relevant spec that
mandated a particular U-A either DOA or be partially implemented.
Neither is a good outcome :)

>   - cookies : the largest ones are those handling an encrypted context so you
>     won't compress them

cookies tend to repeat between transactions on a connection. So the
first encrypted cookie on a connection would not compress well (it will
compress some just because it is expressed in text), but subsequent ones
compress extremely well because of their redundancy. An HTTP/2 with a
lower connection count means more transactions per connection and
therefore greater opportunities for that optimization.

Received on Thursday, 26 January 2012 15:39:09 UTC