W3C home > Mailing lists > Public > ietf-http-wg-old@w3.org > September to December 1995

Re: 'PUT' transaction reconsidered (was Re: two-phase send concerns )

From: Dave Long <dave@navisoft.com>
Date: Wed, 27 Dec 1995 12:27:48 -0800
Message-Id: <199512272027.MAA16426@geek.navisoft.com>
To: "Roy T. Fielding" <fielding@avron.ICS.UCI.EDU>
Cc: Larry Masinter <masinter@parc.xerox.com>, http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com

> I'm becoming more and more convenced that single-exchange PUT is
> unworkable. That is, PUT might have to happen in two steps:

What are the drawbacks to single-exchange PUT if clients are full-duplex? 

The current NaviPress/NaviServer interaction goes like this:

	Press 1.1		Server 1.1
	-----			------
	tries to save file
		-> PUT headers ->
				doesn't have auth
		<- 401 headers <-
	client hasn't noticed
		-> PUT buffer of body ->
				tosses data into bitbucket
		-> PUT buffer of body ->	
			:		"
			:
		-> PUT buffer of body ->
					"
	client finally
	realizes something's wrong,
	while reading response

Now, this isn't too bad at LAN speeds, but it sure is a pain over dialup.

If the server only replies early for failure, then the new logic wouldn't
too bad - put a check for the availability of the response into the
body writing loop:

	Press 2.0		Server 1.1
	-----			------
	tries to save file
		-> PUT headers ->
				doesn't have auth
		<- 401 headers <-
	client hasn't noticed
		-> PUT 0 - 1 buffers worth of body ->
	socket is readable!
	abort, read response...

This gets slightly uglier if the server is allowed to 200 early, in that
we would wind up running the request and response logic in parallel, and
abort the request on a failure response.

As you can see, the NaviServers already support a full-duplex single
exchange (they need to do the bitbucketing anyway, for other reasons),
and I was going to get the logic into NaviPress Real Soon Now.
(Sorry, Roy, a bunch of OEM work came up in the last couple of weeks...)

If y'all have your hearts set on dual exchange PUT, or can tell me
why this single-exchange is too simplistic, I'll abandon efforts on it.

-Dave

I'll probably do the only-failures-early model first, because my experience
has been that an FTP server is the only thing that most of our users ever
get a chance to PUT to, so the NaviServer is at the limits of my world of
interoperability.  If you've got a PUT'able server, send me mail, and I'll
work on keeping us interoperable.
Received on Wednesday, 27 December 1995 12:32:34 EST

This archive was generated by hypermail pre-2.1.9 : Wednesday, 24 September 2003 06:31:38 EDT