[Cosmo-dev] Automated Item submission to a collection

[Brian Moseley]

On 7/26/06, Jared Rhine <jared at wordzoo.com> wrote:
> I wondered aloud on the design list how an Item submitted to a calendar via
> an email address would work technically?  I said there that if I parsed an
> event, server-side as a /Event, I knew how to create an event in a calendar
> via CalDAV or ics PUT.  But how would this work for a generic Chandler Item?
>
> I suggested that maybe the answer is the sharing format.  Is that indeed the
> answer?  Will the operation be something like: Compose an XML/whatever
> single-item file representing the contents of the Item.  Do a PUT with a
> content-type of text/xml to the Collection's URL giving the item a unique
> UUID filename?

in today's world
===========

for a single part email message, i'd suggest:


1) generate a uuid and use that as the name of the resource
2) use the content of the email message as the content of the resource
3) use the content type of the email message as the content type of the resource

for a multipart/alternative message, i'd take the "highest fidelity"
representation of the message (eg if there's text/plain and text/html,
choose the text/html representation) and use that as the resource,
otherwise following the rules above.

for a multipart/related message, i'd consider putting up each part as
a separate resource, potentially rewriting mid/cid urls within them.

this would at least get them into the server. not sure what chandler
would want to do with them once it sees them on the server. it might
be smart enough to recognize resources named with uuids. or perhaps we
could define a webdav property that chandler, cosmo, and the service
know contains the uuid. not sure what other metadata chandler would
require or how it would map arbitrary resources into its data model.

> Alternatives are perhaps CMP instead as Cosmo becomes pretty intimate with
> the Chandler semantics?  Other options?

"morse code" is what the server team has been calling the new sharing
format and protocol, as a working title. when that's documented, and
implementation has begun in chandler and cosmo, the service could
certainly begin to use it as well.