Concerning your earlier message about relative URIs...
I noticed in RFC 2068 the following text:
To allow for transition to absoluteURIs in all requests in future
versions of HTTP, all HTTP/1.1 servers MUST accept the absoluteURI
form in requests, even though HTTP/1.1 clients will only generate
them in requests to proxies.
Is there any way we could use this to our advantage?
At 04:10 AM 6/10/98 +0000, Carl Kugler wrote:
>> At 06:13 PM 6/9/98 -0700, Paul Moore wrote:
>> > I think that originally "printer-uri" was going to be a "virtual"
>> > as you thought. But later (last Fall?) we changed the Model and
>> > document so that the "printer-uri" attribute was required to be
>> > by the client and include in the operation attribute group in the
>> >IPP packet
>> > (which is defined by the application/ipp MIME type). The thinking
>> > was that we wanted the IPP packet and MIME type to be independent
>> > of the transport. So that we could send IPP over any transport,
>> > as SMTP or FTP, for examples.
>> >OK. So we are saying that the URIs IN the protocol are NOT to be used as
>> >transport adresses. They are in effect opaque cookies that the client must
>> >do nothing with except send them back to the printer. They are really
>> >job-name and printer-name. Either that or we explicitly state that these
>> >fields only make sense in an HTTP-enabled environment (they cannot
>> >be mandatory for a universal protocol).
>> No, the URI is actually a URL that is to be interpreted according to
>> "standard" rules for interpreting URLs (not sure if there is a "formal"
>> standard for this). These resource identifiers are not opaque to clients.
>> This does not mean that we are NOT transport independent, it only means we
>> are identifying resources that must be accessed via the transport (scheme)
>> that is specified in the URL. Since we have "modeled" IPP using URI/URL
>> resource identifiers, all transports used by IPP should have a URL scheme
>> defined. I don't think this is a negative constraint BTW. I consider our
>> selection of URL strings as resource identifiers one of the more compact
>> and powerful capabilities of the model (and protocol).
>> The only problem we have identified so far is what happens when "http" is
>> used as the scheme for IPP resources, and between the client and the
>> resource, there is one or more proxies involved. Note, that this is a
>> problem only in the case of HTTP as the transport.
>Another problem is that PRO requires the HTTP Request-URI and the target URI
(embedded in the application/ipp) to be the same (absolute) URIs, but HTTP/1.1
clients aren't allowed to send absolute Request-URIs unless they're talking to
a proxy. And proxies rewrite the Request-URI.
>> For reference purposes, can someone restate the problem (actually the
>> scenario) with proxies that we are trying to address. I think some
>> solutions that have recently hit the list are bordering on "throwing the
>> baby out with the bath water". Any concrete scenario examples would be much
>> appreciated, as I am still on the learning curve with HTTP proxy behavior.