Minutes of HTTP/1.1 editorial teleconference, July 23, 1998

Attending, Larry Masinter, Jim Gettys, Henrik Frystyk, Jeff Mogul, Dave
Kristol, Roy Fielding, Paul Leach.  Minutes by Henrik Frystyk, reviewed
and expanded by Jim Gettys.  Any errors are now my responsibility.

As always, current issues and status are found at
at: http://www.w3.org/Protocols/HTTP/Issues/

Issues discussed (and usually resolved): MUST-MAY-SHOULD, TE-IDENTITY, 
PROXY-DNS, ERRORS, RANGEDELIM, IEBUG, MISTAKES For detail, see the minutes 
below. 

The conference call was made difficult due to the passage of a severe 
thunderstorm through the Boston area (I was in Carlisle, MA, 25 miles 
west of Boston, and Henrik in Boston), which slowed us down and made things 
difficult; the storm forced me to unplug my home machine and deal with 
a portable phone through 7! power-failures.  Consequently, we didn't finish 
going over the issues list today, but hope to finish with another call 
in the next few days.

We plan to discuss in the next (hopefully last) teleconference:
IEBUG, remaining questions in MISTAKES, WARNGEN, and VERSION. 

Draft 04 is in preparation; I plan to submit it the middle of next week 
(well in advance of the IETF ID deadline; I have a family reunion which 
forces my completion well in advance this time).

You should see mail following up from the designated people below on proposed 
language on the working group list in the next two days.

Actions

* h1  MUST-MAY-SHOULD. 
This was classed as technical only as it affects normative language.
There have been no comments on this issue in the list, despite a
call for any comments.  I will apply the edits, doing a final sanity check.

Jim: Editorial

* h2  TE-IDENTITY. Henrik: Add OPTIONAL for how the server handle it. That
is, it is optional for the server to look at it but if it does then it
SHOULD send 406 if it can't respond in the transfer encoding.

* h3  PROXY-DNS.

 a) Paul: Broaden the scope of 504 to contain DNS timeouts as well and
describe this as a minor clarification.

 b) What about authoritative errors in proxies? Too late to handle.  Existing
proxies do at least three different things.  Grist for the HTTP-Extension
working group, who may find additional error codes useful.

 c) Larry: Put in a note for clients saying that to expect in a 400 code
and maybe 500 as well.

* h4  ERRORS. No resolution required - do what the spec says for the
various features.

* h5  RANGEDELIM. Mime-Multipart allows text to be in the area before 
the first boundary. MM says that there should be an extra CRLF so that 
there is an empty line after the boundaries. Dave: Clarify 19.2 that the 
definition of the MM body is defined by RFC 2046 (?). Also make the example 
legal - it misses some CRLFs.

* h6  IEBUG. Believe that if this becomes an important practical problem
then servers will deal with this; it appeared to be an infrequent failure,
on a buggy implementation, without an easy solution not keyed to the
particular browser.

I am slightly nervous about the correctness of this resolution so we'll 
discuss it again in the next phone call, after refreshing our memory on 
exactly what problem 416 solves (no one on the call remembered all the 
details of the problem that caused 416 to be added to the protocol).

* h6a  MISTAKES

a) Section 8.2.3: We believe the current SHOULD should become a MAY; only 
a client ends up having enough information as to whether a retry makes 
sense or not.  Koen's example is a good one, to retry a GET of 1 megabyte 
that failed partially is something only a user agent will know if it wants 
the user control over.  The first sentence also needs editorial work to 
make clear we mean the connection closing before the end of the response 
has been recieved.

b) The consensus was the invalidation was in fact correct, has been a 
requirement for a long time given language existing even in 1945, and 
it would be dangerous to remove it.  At least, that is what I remember
while in the basement with a scared 3 year old near by over the static
in the phone.

c) At this point, the storm arrived in Boston, Henrik went to save his 
machine, and several other attendees had to leave; we'll pick up again 
with Koen's message about section 14.2.  As the storm was mostly over
at this point in Carlisle, this says the storm was about 25 miles across...
Definately the most exciting of any of these teleconferences, at least
for Henrik and me.

				- Jim Gettys

Received on Thursday, 23 July 1998 14:17:00 UTC