Re: Seiwald Q & A -- "GET for EDIT" cookies

David G. Durand (
Tue, 3 Sep 1996 11:42:45 -0400

Larry Masinter <> wrote:
>You know, it might very well be that the right conception of both
>distributed authoring and versioning should be to make them work
>independent of network protocol, by exchange of appropriate
>_messages_. That is, rather than a "CHECKOUT" method, just use "POST"
>with an appropriate "CHECKOUT" method. Rather than returning the data
>as HTML and trying to embed the versioning information inside it,
>return the data in a container (multpart/related comes to mind) that
>has both the versioning information and also the related data.
>I especially like the idea of a "checkin" data object that could work
>with a MAILTO URL as well as an HTTP one.

   I don't have a problem with this, except that we have HTTP PUT, and the
versioning stuff must work with it (to my mind, PUT is un-useful without
versioning, but that is (perhaps) only my problem). If we move all the
versioning stuff into a special content-type that must be processed,
versioning updates will be implemented differently from non-versioning
updates. This will reduce the ability of versioning to work with legacy
authoring environments, and _require_ special work for versioning to be
supported at all.

   I think this violates some of the requirements, unless we remove PUT,
define POST as "form-only" and make all updates use a content-type (w/
simple options that allow versioning).

   I liked the multipart/HTTP you suggested for transactions, but that's a
CM and not a versioning issue. I think that we agree that CM support will
not work well will legacy software, except insofar as we can support the
update of single resources.

  -- David

David Durand         |
Boston University Computer Science          | Dynamic Diagrams    |