Re: About that Host: header....

Although I agree with you (John) that the medium-term viability
of the Internet may depend on getting Host: widely implemented,
I think you are fantasizing if you think that any punitive
mechanism will make this happen any faster.  I would count
"changing the version number" or "requiring full URLs in HTTP/1.1"
as punitive mechanisms, since the penalty for not conforming
to these will be non-interoperation.

If we try to impose these requirements, the major vendors won't
go along (because their customers will feel the pain immediately,
and will not see any immediate gain).  So instead of solving
what is by any measure NOT a short-term problem, we will destroy
the rather fragile cooperative environment that allows us to
have an HTTP standard at all.

No matter how important Host: is to the Internet, trying to
impose it in such a way that causes immediate pain WILL NOT WORK.

I submit that
	(1) As long as the major browser and server vendors
	are cooperating with the HTTP standardization process,
	we should expect that the vast majority of HTTP/1.1
	implementations will do Host: according to the current
	proposed spec.; i.e., we ought to trust people for now.

	(2) There will always be a few browsers out there that
	don't send Host:; there is absolutely nothing we can
	do about it.  So we should be working on getting HTTP/1.1
	out as soon as possible, and with as much cooperation
	from vendors, with the intent of reducing the number
	of obsolete browsers as quickly as possible.  But we
	shouldn't fool ourselves that this will eliminate
	vanity host addresses any time soon (but neither will
	trying to impose a standard that nobody implements).

	(3) The best way to get people to change their behavior
	is to provide an incentive based on self-interest.
	Van Jacobson's slow-start mechanisms became popular
	not just because they made the Internet work better,
	but because they also made individual connections work
	better, and so even greedy people wanted Van's code.
	So if we want to increase the rate of adoption for
	Host:, perhaps we should find some incentive to tie
	to it.

I don't have a great suggestion for #3, but I suggest that
this is the direction we ought to be thinking in.

I think we all agree that, in retrospect, the original HTTP
design was faulty in not requiring full URLs.  But forming
a circular firing squad is no solution.

-Jeff

Received on Wednesday, 20 March 1996 17:47:32 UTC