Questions about HTTP/V1.0 doc dated 14-Jul-93

Fred Williams (fwilliam@ccs.carleton.ca)
Wed, 11 Aug 93 5:54:21 EDT


RE: Questions -- Hypertext Transfer Protocol Internet Draft 14-Jul-93

I have just read the document and a number of questions have come up concerning
the actual implementation.

1) The ProtocolVersion stated in 1. is shown to be HTTP/V1.0 and also in 2.
however in 1.1 Protocol Version it is stated 'only HTRQ is defined'.
I assume HTRQ, based on an earlier document, is correct and
is the ProtocolVersion is really HTRQ/V1.0?

2) How does one terminate HTRQ header if there is no data to follow? I
realize that a blank line separates header from data so does one still
include a blank line if there is only header information?

3) An earlier document on HTTP describes the TEXTSEARCH method to request a
search to be done utilizing search information contained in the data
section. Since the data section is in MIME this would allow complex
searches to be done and searches done on information other than ASCII text.
The latest document seems to degrade method TEXTSEARCH to only a server
response to indicate that searching is possible. One then uses the
GET method ( ie ?xxx+yyy ... ) to perform the actual search. This does
now restrict the search capabilities to only ASCII and simple searches.
(ie 'very' limited pattern matching and no boolean operators ). Is it
now intended to keep TEXTSEARCH only as a response ( ie duplicating
[ISINDEX] functionality) or to return its functionality as a request
method.

4) What does HTRQ stand for? My best guess to date is HyperText ReQuest

5) The abstract in the document states that 'It is a generic
stateless object-oriented protocol'. however it is suggested in 1.7.6
Referer: that the information send could be used by the server to
generate lists of back-links to documents. This functionality would
not make the server `stateless'. Is this what is intended?

6) The CHECKIN and CHECKOUT methods do not have an associated server
response. This would be required so that the server could return a
`lock indentifier` that would have to be used to unlock the object at
a later date. This would then maintain the 'statelessness' of the
server. Also how does the client set the lock timeout or is it up to
the server to assign an arbitrary timeout to the lock?

With Thanks

Fred Williams
fwilliam@ccs.carleton.ca