On 13/2/06 5:25 PM, "Tim Bray" <[EMAIL PROTECTED]> wrote:

> I'm in favor of simplification, but there's one scenario that seems
> harder to me without media collections.  Suppose I'm a cameraphone.
> I want to create a post containing one picture and a few words
> wrapped around that.  How do I do that using APP?

Could it not POST a Multipart/Related (atom entry + image binary), which
returns a HTTP Location header for the Atom Entry, in which is referenced
where the image binary is located (either via atom:link or [EMAIL PROTECTED],
whichever is decided).

Would the cameraphone need to be able to retrieve the image once uploaded,
or to submit an edited (eg. cropped or rotated) version? Would the camera
phone need to be able to retrieve, edit, and resubmit the "few words"
involved? 

The absolute simplest would have the cameraphone act as a one-way sender,
not requiring any capability to GET and parse XML, only needing to remember
the Location header and associate that with the image+words, such that it
could then do a blind PUT of a Multipart/Related (atom entry + image binary)
to that location. 

Absolute simplest wouldn't even require the cameraphone to understand XML or
MIME at all ... just concatenate some strings together (shock! horror!). The
server shouldn't care if the multipart boundary IDs on different post
requests look exactly the same, for e.g, right?

The XML+MIME overhead would be trivial compared to the binary payload,
right? What is the minimal XML required to post? If a client can leave out a
bunch of stuff it can assume the server will generate, I count about 150
bytes (obviously not legal Atom Syndication Format, the server would need to
fill in the missing bits).

What wouldn't work?

e. 

Reply via email to