Re: [cmis] joining the TC, introduction, HTTP thoughts

["David Nuescheler" <david@day.com>]

Hi Al,

Thanks a lot for your mail.

Let me comment on a couple of points.

> I also believe that CMIS is (or can be mapped) to a subset of JCR. So it
> would be interesting for the JCR EG to define a profile/level that can exist
> directly on top of a CMIS-compliant repository. That would give Java
> developers alternatives to Abdera or JAX-WS APIs. I am all for that.
That's great to hear.

While I agree that a JCR mapping would give Java developers
more options, I would really like to stress that I do not see
that at the same level as the Apache Abdera Project.
JCR is already an integral part of Abdera since v0.4 (see release
notes [1]) and
since Abdera and Jackrabbit are sister projects at Apache the two
communities interact well based on the fact that they operate under
the same licensing and intellectual property framework within the ASF.

The additional value that the JCR / CMIS mapping would bring to the
table is that helps us understand where the models are different and
why the decision that we took in JCR are now different in CMIS.
While of course it is perfectly fine to differ from any existing standards
(not just JCR) I think the TC should do that consciously rather
than "by accident".

In my mind this perfectly aligns with the charter of creating liaisons
to other respective standards like HTTP, WebDAV, JCR and hopefully
AtomPub.

There are a number of things that in hindsight we would have done
differently in all of these specification and there is definitely no need
to make the same mistakes twice, but for a large part the experts that
defined the respective specifications at the time, had good reasons
to do things a certain way, which means that the rationale at the time
may also apply to CMIS.

> I have also had interesting discussions with David N about Jackrabbit
> exposing CMIS. As Jackrabbit is a repository, it makes sense to encourage
> that group to expose a CMIS interface as well. I think that will be very
> beneficial for the effort. We already have a repository that supports the
> current CMIS model and JCR - Alfresco.

I would assume that the Alfresco implementation maps CMIS to its internal
proprietary interface not to JCR (maybe Dave or John could comment on
that, I may be wrong since I try to look at GPL'd code as little as possible
for obvious reasons).

While mapping to a proprietary API is of course a perfectly good solution
for any particular vendors implementation it does not really help us find the
mapping and differences between JCR and CMIS which I think would be
desirable, as mentioned above.

I do not think that it is something that this TC would have to attend
to but it is something that is going to be developed eventually as an open
community effort in the Apache Software Foundation anyway, so I will keep
the group posted on any potentially relevant findings.

regards,
david

[1] http://abdera.apache.org/