W3C home > Mailing lists > Public > ietf-http-wg@w3.org > January to March 2013

Re: Delta Compression and UTF-8 Header Values

From: Willy Tarreau <w@1wt.eu>
Date: Sun, 10 Feb 2013 12:05:14 +0100
To: Poul-Henning Kamp <phk@phk.freebsd.dk>
Cc: Frédéric Kayser <f.kayser@free.fr>, ietf-http-wg@w3.org
Message-ID: <20130210110514.GR8712@1wt.eu>
On Sun, Feb 10, 2013 at 10:29:12AM +0000, Poul-Henning Kamp wrote:
> Content-Type: text/plain; charset=ISO-8859-1
> --------
> In message <20130210101248.GQ8712@1wt.eu>, Willy Tarreau writes:
> >On Sun, Feb 10, 2013 at 09:38:03AM +0000, Poul-Henning Kamp wrote:
> >> The only two places which care about the character-set of the URL,
> >> is the ultimate client and the ultimate server, to everybody else,
> >> it is just a sequence of opaque bits, which they must treat as a
> >> indivisible unit.
> >
> >It's not that much opaque when your "HTTP router" has to be able to
> >match part of that URL to decide where to route the requests. 
> 
> First of all, the only non-semantic criteria you can route HTTP
> requests on are the Host: header.  (Which we just pass to DNS,
> so we don't care if it is UTF-8 or not).
> 
> As soon as you look at the URL, you dive into semantics, and you
> had better have an agreement with the content-provider about what
> those semantics (including char-set) and routing-criteria should be.

The content provider is the same as the one which sets the routing rules.
The 2 typical uses are :
  - send static objects to a dedicated farm
  - bypass caches for definitely non-cacheable objects

I would love to see only configurations where Host is the only
criterion but real world differs from the ideal one.

Willy
Received on Sunday, 10 February 2013 11:05:45 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Sunday, 10 February 2013 11:05:47 GMT