- From: Bob Jernigan <jern@spaceaix.jhuapl.edu>
- Date: Wed, 3 Apr 1996 15:35:06 -0500 (EST)
- To: Paul Leach <paulle@microsoft.com>
- Cc: http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com, ddubois@spyglass.com
Paul Leach wrote: > > > > >---------- > >From: Daniel DuBois[SMTP:ddubois@spyglass.com] > >Subject: Re: 505 HTTP Version Not Supported > > > >At 01:42 PM 4/3/96 -0500, Bob Jernigan wrote: > >>I.e., "if I don't need it, no-one does." > > > >I don't think it's a matter of me shrugging off the little people. > >We've > >been moving a long time towards increased HTTP complexity. We're now > >at the > >point where to be compliant with the latest HTTP there are headers you > >are > >*required* to send. The issue of how virtual servers are supposed to > >handle 0.9 requests can be a sticky one (as long as 1.0 and 0.9 are > >around > >will vendors continue to suck up multiple IP addresses for one > >machine?). > > > >IMO, this pseudo-backward compatibilty can only remain feasible for so > >long. > >I just think the world (wide web) would be a better place if no one > >made 0.9 > >requests anymore. And I'll probably feel the same way about 1.0 > >requests > >soon enough. > > > >>But 0.9 is useful and will continue > >>to be useful in certain circumstances. About half my hits are 0.9 but > >>they do return <1% of the data. It all depends on what you think http > >>is useful for. > > <soapbox> > I think Dan is being too kind. My reading of how people in charge of big > chunks of the Interent backbone feel is that if we don't expunge 0.9 > _and_ 1.0 ASAP then the whole net will collapse. So, while it may be > convenient to not have to change, and the old way may deliver value to > those using it, they are creating disvalue for everyone else. > > The minimum pieces of 1.1 that have to be implemented by clients and > servers to "save the Internet" aren't that hard: send Host:, and use > persistent connections. (Proxy caches have additional requirements.) The > people who have "20 line HTTP clients and 200 line HTTP servers" in PERL > might see their code grow by 20 lines in order to do this. Real clients > and servers might add 100-1000 lines (mostly for persistent > connections). In return, the Internet will be there for us all to use. > </soapbox> > > Paul > Then perhaps it's time to be explicit about retiring 0.9. I find it useful in my intranet environment but I see no reason this should burden the Internet. We could say that "If you use 0.9 in an Intranet environment, then this environment should be shielded from the Internet." I can't believe I'm the only one still using 0.9, but if so, I'll happily withdraw any request to preserve 0.9 and stuff them in a tiny server behind my firewall and drop any claim about HTTP. The same thing might apply to anyone who wants to stay with HTTP/1.0. bob
Received on Wednesday, 3 April 1996 12:41:14 UTC