Re: encoding and other h2 extensions

On 22 November 2014 03:25, Poul-Henning Kamp <phk@phk.freebsd.dk> wrote:

> ​​
>
> We don't need the choice between Compress+Gzip+Zlib, just pick
> one of them and the world will be a better place.
> ​​
>
> ​<...>
>
>
> If somebody comes up with a compression which is significantly
> better than gzip, for this purpose, then *that* could be added
> to the list.
>
>
​So I guess your preference is gzip?​

​​
>
> There are no credible schenarios under which a HTTP/2 implementation
> cannot implement *any* of those three, but it is utterly stupid
> to require them to implement all three, just to be sure to interop.
>
> ​
To my mind, the initial registry is of secondary importance to getting the
machinery right in the first place. If the extension mechanism itself
works, and issues like fragmentation are significantly addressed, *then*
I'd want to focus on seeding the registry (in a way that will promote
adoption and interop). This list was pulled straight from RFC 7230 S4.2,
plus a little splash of paint, because that involved the least thought on
my part.

I have no problem with reducing it to just gzip, say. I've also had a
request for LZ4, which is a damn sight faster than LZ77, if a bit weaker on
the compression ratio. The problem with that one is that it doesn't have
gzip's ubiquity, and I don't know how to reference things that don't have
solid specs (at least more solid than a blog post[1]).

​But as you say, the encodings themselves ​
​
​are​ almost trivial; I want to get the extension part right first.




​[1]: http://fastcompression.blogspot.com.au/2011/05/lz4-explained.html​


-- 
  Matthew Kerwin
  http://matthew.kerwin.net.au/