Re: About that Host: header....

>In any case -- whether the host is in the URL or the Host: header is a
>mere matter of taste.  One of them breaks things, the other doesn't --
>so which one do we pick?

I wouldn't say "mere" taste. If one has one wart on the spec thats OK.
Ten or twenty and the thing will start to get raqged. As I see it we have
two issues:

1) Backwards compatiblility with installed server base.

2) Desire to have 1.2 be clean.


Can we enforce this as follows:

1) A http server must accept either a full URI _or_ a host header but
	not both.

2) A http 1.2 client will be required to use a full URI if 1.2 is reported.


Note that a 1.2 server would be required to accept a host header even
though a 1.2 client would no longer generate one. In that sense the
strict forwards compatibility of the 1.x series would be preserved while
ensuring that we would move to using only full URIs in a reasonably rapid
timeframe. 

Jims suggestion that a 1.1 server produce an error if a message claims to 
be 1.1 but is not also sounds a good move.


I think that John has a point about wanting to avoid carbunkles but
we can't really expect to introduce a change that will cause undue numbers
of breakages.

		Phill

Received on Monday, 18 March 1996 22:56:24 UTC