Re: An MGET proposal for HTTP

Dave Kristol (dmk@allegra.att.com)
Mon, 31 Oct 94 09:58:32 EST


I would like to add my thanks to John Franks for setting forth an MGET
proposal. Straw proposals are much easier to debate than are the
hypotheticals that have been hand-waved at here. I agree with his
objectives (not repeated here). However, like some other writers, I
have some quibbles with/questions about the proposal.

john@math.nwu.edu (John Franks) writes:
> Proposal for an HTTP MGET Method
[Design Objectives omitted]

1) Consider authentication. Suppose the server requires authentication
for one or more of the URL's in the MGET. What would the
interaction(s) between client and server look like? Consider,
particularly, what happens if some of the elements of the MGET fall
into different protection realms and thus require different methods of
authentication (or different username/passwords, to keep the discussion
simple).

2) I would like to see a definition of Content-type-encoding: packet.
Specifically, which bytes are included in the byte count? CRLF's? Is
the data uuencoded? Or are the bytes assumed to be binary data? I'm
asking for clarity because, if clients and servers get out of sync with
the counts on the MGET response, the result at the client side will be
a real mess.