Re: Entity too Large solutions

On Thu, Nov 14, 2002 at 11:28:17AM +0100, [email protected] wrote:
> is someone allready looking into the extensive amount of headers that 
> kannel generates? I can imagine that by concatenating all accept content 
> types on one line (and the same for character codings) the web-servers can 
> be much more happy.

That depends.  Some web servers get very unhappy if header lines are
very long.  In order to help, such concatenation would probably have
to be "smart", for example by concatenating as much as possible while
staying under a line length limit.

> I am however not too sure whether such a solution is acceptable 
> considering the encoding used at the WSP layer. The WSP specification 
> neither requires not forbids the gateway to concatenate all acceptable 
> contenttypes into one Accept header.

The HTTP/1.1 specification (rfc2616, section 4.2) requires support for this.
In fact, it lists _not_ concatentating (that is, having multiple headers
with the same name) as a special case.

If I remember correctly, the WSP specification never really says whether
multiple identical header fields in the bytecode are supposed to be one
header line or many.  I think it's up to the implementation to decide.

When encoding response headers, it's pretty clear that a single header must
be encoded as multiple header fields, because the bytecode format doesn't
have room to do it any other way.  So I think there's at least the intent
that header splitting should not matter.

Richard Braakman